<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Archiving and Interchange DTD v1.0 20120330//EN" "JATS-archivearticle1.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink">
  <front>
    <journal-meta>
      <journal-title-group>
        <journal-title>ISO</journal-title>
      </journal-title-group>
      <issn pub-type="ppub">1613-0073</issn>
    </journal-meta>
    <article-meta>
      <title-group>
        <article-title>What Generative Artificial Intelligence Means for Terminological Definitions</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <string-name>Antonio San Martín</string-name>
          <email>antonio.san.martin.pizarro@uqtr.ca</email>
          <xref ref-type="aff" rid="aff0">0</xref>
        </contrib>
        <aff id="aff0">
          <label>0</label>
          <institution>University of Quebec in Trois-Rivières</institution>
          ,
          <addr-line>3351, boulevard des Forges, Trois-Rivières, Quebec G8Z 4M3</addr-line>
          ,
          <country country="CA">Canada</country>
        </aff>
      </contrib-group>
      <pub-date>
        <year>2024</year>
      </pub-date>
      <volume>704</volume>
      <issue>2022</issue>
      <fpage>27</fpage>
      <lpage>28</lpage>
      <abstract>
        <p>This paper examines the impact of Generative Artificial Intelligence (GenAI) tools like ChatGPT on the creation and consumption of terminological definitions. From the terminologist's point of view, the strategic use of GenAI tools can streamline the process of crafting definitions, reducing both time and efort, while potentially enhancing quality. GenAI tools enable AI-assisted terminography, notably post-editing terminography, where the machine produces a definition that the terminologist then corrects or refines. However, the potential of GenAI tools to fulfill all the terminological needs of a user, including term definitions, challenges the very existence of terminological definitions and resources as we know them. Unlike terminological definitions, GenAI tools can describe the knowledge activated by a term in a specific context. However, a main drawback of these tools is that their output can contain errors. For this reason, users requiring reliability will likely still resort to terminological resources for definitions. Nevertheless, with the inevitable integration of AI into terminology work, the distinction between human-created and AI-created content will become increasingly blurred.</p>
      </abstract>
      <kwd-group>
        <kwd>terminological definition</kwd>
        <kwd>Generative Artificial Intelligence</kwd>
        <kwd>ChatGPT</kwd>
        <kwd>AI-assisted terminography</kwd>
      </kwd-group>
    </article-meta>
  </front>
  <body>
    <sec id="sec-1">
      <title>-</title>
      <p>CEUR
ceur-ws.org</p>
    </sec>
    <sec id="sec-2">
      <title>1. Introduction</title>
      <p>
        Definitions are an essential element in terminological resources since they explain the
conceptual content that a given term can convey. However, the creation of terminological definitions,
especially the selection of the information to include in them, is very time-consuming.
According to the classical approach1, the characteristics featured in a definition (i.e., the necessary
and suficient characteristics of the defined concept) are universal and context-independent.
However, they are most often impossible to determine objectively [
        <xref ref-type="bibr" rid="ref41 ref8">8</xref>
        ], and even when possible,
this approach yields less useful definitions for non-experts.
      </p>
      <p>
        According to the Flexible Terminological Definition Approach [
        <xref ref-type="bibr" rid="ref42 ref9">9</xref>
        ], accounting for the role
of context in meaning construction is necessary to overcome the drawbacks of the classical
approach and create definitions that fulfill the user’s needs. From a cognitive linguistics
perspective, terms, like any other lexical unit, do not possess meaning in themselves but are
simply access points to large repositories of knowledge [
        <xref ref-type="bibr" rid="ref10 ref43">10</xref>
        ]. It is context, understood broadly as
3rd International Conference on Multilingual digital terminology today. Design, representation formats and management
any factor that can afect interpretation [
        <xref ref-type="bibr" rid="ref11 ref44">11</xref>
        ], that determines which segment of this knowledge
(i.e., which meaning) is activated in each usage event.
      </p>
      <p>
        All the knowledge that a term can invoke is its semantic potential [
        <xref ref-type="bibr" rid="ref12">12</xref>
        ], [
        <xref ref-type="bibr" rid="ref13">13</xref>
        ]. For instance,
while the semantic potential of light fuel oil includes all knowledge that it can activate in any
context, the knowledge it conveys in a specific context, like in Article 20 of the Canadian
Greenhouse Gas Pollution Pricing Act, is an instance of meaning. Describing a term’s semantic
potential in a definition is not feasible because of its vastness. Moreover, explaining meanings is
not the goal of definitions, as meanings are transient [
        <xref ref-type="bibr" rid="ref14">14</xref>
        ]. When terminologists craft definitions,
they must select the most relevant information from the term’s huge semantic potential based on
contextual constraints (linguistic, thematic, cultural, ideological, geographical, and chronological
[
        <xref ref-type="bibr" rid="ref15">15</xref>
        ]) and functional constraints (user needs, resource characteristics, etc.). Applying contextual
constraints to a term’s semantic potential results in a specific subset known as a premeaning,
which is what a terminological definition describes.
      </p>
      <p>
        A premeaning is a conceptualization unit halfway between a term’s semantic potential and
meanings in particular usage events [
        <xref ref-type="bibr" rid="ref16">16</xref>
        ]. An example of a premeaning is the knowledge
activated by eutrophication in the domain of Wastewater Treatment in a Jamaican geographical
context. This contrasts with its broader semantic potential and with all its possible narrower
meanings (e.g., its meaning in a tweet published by the UN Environment Programme on January
31, 2021).
      </p>
      <p>If crafting traditional definitions is labor-intensive, the consideration of contextual and
functional constraints makes the task even more time-consuming. This is an important barrier
to the creation of flexible terminological definitions. Generative Artificial Intelligence (GenAI)
tools, especially those powered by Large Language Models (LLMs) such as ChatGPT2, can
remove these barriers by reducing the time and efort required to create them. However, the
impact of GenAI can extend well beyond this, as it can profoundly transform the methods and
purposes underlying the creation and consumption of terminological definitions.</p>
      <p>
        With varying levels of reliability, ChatGPT can fulfill all the terminological needs of a user
(interlingual equivalents, denominative variants, collocations, examples of use, pronunciation,
etymology, etc.), including term definitions. ChatGPT can also assist terminologists in their work.
Given this enormous potential, we can wonder, as [
        <xref ref-type="bibr" rid="ref17">17</xref>
        ] does for Lexicography, whether ChatGPT
will replace terminological resources, terminologists, and the tools used for terminology work.
Still another question is if ChatGPT will replace terminological definitions, the terminologists
that write them, and the tools used for definition writing. While it is currently not possible
to answer these questions because of the quickly evolving nature of GenAI, in the following
sections, we explain some factors that may influence the future of terminological definitions in
the face of AI.
2Throughout the article, we refer to ChatGPT to represent all LLM-based tools, as it is the most popular at the time
of writing. However, most observations presented here also apply to other GenAI chatbots (i.e. Google Gemini or
Anthropic’s Claude). It is important to note that while terminologists currently mainly use LLMs through chatbots,
GenAI integration into various applications, such as corpus analysis tools and terminology database managers, is
to be expected.
      </p>
    </sec>
    <sec id="sec-3">
      <title>2. GenAI and the future of terminological definitions</title>
      <p>A query to ChatGPT has significant advantages over consulting a definition in a terminological
resource. The main one is that while a definition can only describe a premeaning, ChatGPT can
describe meaning, i.e. the knowledge activated by a term in a specific usage event. A user can
ask ChatGPT to explain the meaning of a term in a specific context (Example 1) 3. ChatGPT can
also tailor its responses based on the user’s needs. Users can request term definitions within
particular contexts — such as a text, an image, or certain situation — and receive answers suited
to their existing knowledge (Examples 2-4).</p>
      <p>The chat interface enables the user to ask follow-up questions to seek clarifications, request
examples, etc. This interactive approach enhances the user’s understanding of terms in a way
that definitions cannot. Additionally, despite the limitations in how ChatGPT currently accesses
online information, it can ofer up-to-date information, unlike terminological definitions that
may become obsolete over time. Finally, ChatGPT stands out for its user-friendliness, and it
can answer questions on virtually any subject, obviating the need for users to preliminarily
identify the relevant resource for their needs.</p>
      <p>
        However, from the perspective of a user needing a definition, ChatGPT has certain drawbacks.
More specifically, its responses may contain factual, logical, and linguistic errors, among other
types [
        <xref ref-type="bibr" rid="ref18">18</xref>
        ]4. They can also be biased and discriminatory, and ChatGPT can even hallucinate,
i.e., fabricate incorrect information. Another problem related to defining is that ChatGPT has
dificulty with handling sense splitting and lumping [
        <xref ref-type="bibr" rid="ref20">20</xref>
        ]. This can lead to the omission of
senses or the unnecessary subdivision of a single sense when prompted to create a definition.
However, when provided with suficient context, ChatGPT generally manages to select the most
relevant one.
      </p>
      <p>
        Furthermore, ChatGPT generates diferent responses to identical queries, which is a challenge
in situations requiring a single unchanging definition. Lastly, its environmental footprint [
        <xref ref-type="bibr" rid="ref21">21</xref>
        ],
as well as the ethical and legal issues arising from its training on copyrighted materials [
        <xref ref-type="bibr" rid="ref22">22</xref>
        ],
can be major drawbacks for some users.
      </p>
      <p>In the face of these limitations, the main advantage of a terminological definition inserted
in a resource lies in its reliability. In fact, in our view, reliability is the main reason why
terminological resources, or at least some of them, will continue to be necessary despite GenAI.
Nevertheless, ChatGPT’s unreliability is likely to pose a challenge only for some users in
particular contexts. For instance, a translator reading a blog post for pleasure may resort to
ChatGPT to understand an unfamiliar term because of its advantages and the low stakes of
potential errors. However, for professional tasks, they are more likely to directly consult a
terminological resource. Some will also probably turn to ChatGPT first and subsequently verify
the information with reliable sources.</p>
      <p>
        Some definitions have a higher chance of surviving GenAI than others. One case is the
definitions with a prescriptive or standardizing orientation or those aiming to explain how terms
are to be interpreted in specific contexts. These definitions can be found in standards, patents,
3Examples of ChatGPT prompts and responses are included in the appendix.
4The problem of erroneous responses is more pronounced when interacting with ChatGPT in languages other than
English [
        <xref ref-type="bibr" rid="ref19">19</xref>
        ]. Responses in other languages may also exhibit interference from English, especially in language-related
queries [
        <xref ref-type="bibr" rid="ref20">20</xref>
        ].
legislation, or termbases reflecting an organization’s own terminology. These definitions will
not probably be replaced by AI because they intend to reflect a consensus among human beings.
Another case is those definitions provided in some publications, especially pedagogical ones,
to explain terms mentioned throughout the main text. They are typically found in footnotes,
information boxes, or glossaries found at the end of the publication. As in-text definitions that
authors provide when a term is mentioned for the first time, these definitions can survive AI
because they have the advantage of being accessible to the user exactly when they are needed.
      </p>
      <p>As for the potential extinction of the professional terminologists due to AI advancements, it is
noteworthy that the above-mentioned definitions that are less susceptible to being replaced by AI
are often not created by terminologists. However, as already discussed, terminological resources
that include definitions created by terminologists are not likely to disappear. Nevertheless, their
number may decrease as they become less commercially viable or lose public funding. It is
believed that AI could fulfill their role. Additionally, as discussed in Section 3, the boundary
between human-created terminological resources and definitions and those generated by AI
will become increasingly blurred.</p>
    </sec>
    <sec id="sec-4">
      <title>3. AI-assisted terminography</title>
      <p>
        Given the expected pervasiveness of AI, terminologists will benefit from varying degrees of
assistance from AI technologies, in what can be termed AI-assisted terminography. This would
include all forms of terminographic work where AI supports the terminologist. A particularly
relevant type, especially for terminological definition writing, is post-editing terminography
(an adaptation of the term post-editing lexicography [
        <xref ref-type="bibr" rid="ref23">23</xref>
        ]) in which the machine produces a
terminographic item (e.g., a definition) and then the terminologist corrects or refines it.
      </p>
      <sec id="sec-4-1">
        <title>3.1. Post-editing terminography</title>
        <p>In a post-editing terminography context, the strategic use of ChatGPT can enable terminologists
to create definitions faster without compromising quality. This eficiency will allow
terminological resources to remain up-to-date and comprehensive, and at the same time, keep pace with
the swift emergence of new terms and concepts. However, this is not without its challenges.</p>
        <p>Firstly, the potential for errors and biases requires careful verification, which can be as
labor-intensive as crafting the definition from scratch. Productivity pressures could lead, in
some cases, to erroneous definitions generated by ChatGPT being published without proper
validation.</p>
        <p>
          Furthermore, ChatGPT’s responses occasionally constitute plagiarism [
          <xref ref-type="bibr" rid="ref24">24</xref>
          ] or reproduce
copyrighted content verbatim [
          <xref ref-type="bibr" rid="ref25">25</xref>
          ]. Unless ChatGPT bases its answer on an Internet search,
it cannot typically identify its sources, and, when prompted, it may generate fictitious ones.
These issues present legal and ethical concerns and complicate the task of assessing the
reliability of the information provided by ChatGPT. This contrasts with corpus-based post-editing
lexicography/terminography tools that enable users to consult the corresponding concordances,
like the integration between Sketch Engine and Lexonomy [
          <xref ref-type="bibr" rid="ref23">23</xref>
          ].
        </p>
        <p>In AI-powered post-editing terminography, terminologists should be aware of the types of
potential errors and remain vigilant to identify and correct them as needed. Furthermore, the
post-editing efort that ChatGPT-generated definitions require can be reduced by employing
optimized prompts5, where terminologists specify as much as possible the characteristics of the
definition.</p>
        <p>A definition can be obtained by ChatGPT by prompting simply ”Define” followed by the term
to be defined. However, this approach tends to yield a lengthy encyclopedic answer (Example
5). Using “Write a terminological definition of…” normally yields a more post-editable result
(Example 6). However, ChatGPT does not always follow the basic rules of definition writing,
such as not repeating the defined term at the beginning of the definition. This can be solved by
asking ChatGPT not to mention the defined term at the beginning of the definition (Example 7) 6.</p>
        <p>If no contextual constraints are provided, ChatGPT typically defines terms by encompassing
their most common conceptualizations across relevant knowledge domains. Furthermore,
because of the nature of its training data, these definitions often tend to reflect a Western-centric
perspective. Specifying in the prompt as many contextual constraints as possible as well as
functional ones (target user of the definition, resource in which it will be inserted, etc.) can
reduce the subsequent post-editing efort (Example 8).</p>
        <p>When dealing with polysemic terms, it can be strategic to instruct ChatGPT to provide multiple
definitions to define the diferent associated concepts. Without clear directives, ChatGPT might
either omit certain concepts or merge them into a single definition. However, even with these
instructions, terminologists should be aware that ChatGPT might omit relevant concepts or
split concepts unnecessarily (Example 9).</p>
        <p>
          ChatGPT is also capable of following definitional templates, which indicate the type of
definitional features that the definitions of terms in the same category should include [
          <xref ref-type="bibr" rid="ref27">27</xref>
          ]
(Example 10). It is possible to provide ChatGPT with either empty or filled-in templates. For
example, a template can indicate that in the case of antihypertensive drugs, the definitions
should contain the route of administration. The terminologist can specify which are the routes
of administration of each drug to ChatGPT or let ChatGPT fill this information in by itself based
on its knowledge. Obviously, the more information the terminologist provides, the less risk of
errors and hallucinations.
        </p>
        <p>It should be noted that post-editing terminological definitions is currently only viable in
highresource languages, as the quality of ChatGPT-generated definitions in less-resourced languages
may not be adequate for post-editing. Likewise, in the case of new terms or concepts not found
in the ChatGPT training data, the result could also be suboptimal, even with ChatGPT’s ability
to base responses on web searches. Consequently, other forms of AI-assisted terminography
are essential, which in turn can also be useful for validating AI-generated definitions.</p>
      </sec>
      <sec id="sec-4-2">
        <title>3.2. Other forms of AI-assisted terminography</title>
        <p>To inform definition writing, terminologists consult reference work and specialized texts. Given
that terminologists must often deal with knowledge domains that are unfamiliar to them,
ChatGPT can help solve all kinds of notional doubts (Examples 11-12).</p>
        <p>
          Another useful method for definition writing is the analysis of existing terminological
definitions of the term to be defined (when available). It follows that segmenting and contrasting
5[
          <xref ref-type="bibr" rid="ref26">26</xref>
          ] is an example of prompt optimization lexicographical definition writing.
6Asking ChatGPT to produce a lexicographic definition tends to yield a complete lexicographic entry.
definitions yield structured semantic information [
          <xref ref-type="bibr" rid="ref28">28</xref>
          ], and repeated features are likely to be
relevant [
          <xref ref-type="bibr" rid="ref29">29</xref>
          ]. While ChatGPT is currently not more helpful than a search engine to collect
existing definitions, it is certainly helpful in analyzing them. After feeding ChatGPT a list
of definitions of the same term with their source, ChatGPT can be asked to extract a list of
semantic traits from them (Example 13). Once the list has been generated, the user can ask
ChatGPT to perform a number of operations, such as checking the accuracy of each semantic
feature (Example 14), or ranking features according to their relevance in relation to certain
contextual constraints (Example 15). ChatGPT can assist in identifying gaps in the definitions
by highlighting any missing information (Example 16). ChatGPT can also write a post-editable
definition using the list of semantic traits and justify it based on the list (Example 17).
        </p>
        <p>ChatGPT can also support definition writing by assisting with corpus analysis. Manually
analyzing concordance lines to find definitional information is extremely time-consuming.
ChatGPT can be fed concordance lines and asked to extract semantic information for definitional
purposes. ChatGPT can help analyze concordance lines in search of semantic information related
to a term to be defined (Example 18). However, due to the size limitations for ChatGPT prompts,
if too many concordance lines are available for the term to be defined, the terminologist must
pre-select relevant ones before submitting them to ChatGPT, akin to how they choose specific
concordance lines for manual analysis in traditional corpus analysis.</p>
        <p>
          An eficient way of selecting concordance lines to be analyzed is word sketches, which are
automatic corpus-derived summaries of a word’s behavior created by Sketch Engine [
          <xref ref-type="bibr" rid="ref30">30</xref>
          ]. They
list words that appear in the corpus in a given relationship with the search word and give access
to the corresponding concordance lines. Default Sketch Engine word sketches mostly represent
linguistic relations (verb-object, modifiers, etc.). Although particularly useful for collocation
analysis, they can also be used for semantic analysis. For instance, by choosing concordances
with word sketches, we can ask ChatGPT to analyze the semantic information contained in a
list of concordance lines where the target term is the object of the sentence (Example 19).
        </p>
        <p>
          A special type of word sketches that are useful for specialized semantic analysis are semantic
ones [
          <xref ref-type="bibr" rid="ref31">31</xref>
          ], [
          <xref ref-type="bibr" rid="ref32">32</xref>
          ] because they extract the other terms that hold a semantic relation (hypernymy,
meronymy, cause, etc.) with the search term and give access to the corresponding concordance
lines. For instance, it is thus possible to ask ChatGPT to extract all the semantic information
contained in a list of concordance lines where the target term holds a causal relation with other
terms (Example 20).
        </p>
        <p>ChatGPT can generate post-editable definitions using concordance lines as source material
and justify its responses based on them (Example 21). However, manually copying and pasting
concordances from corpus tools into ChatGPT can be time-consuming and ineficient. The
anticipated integration of AI into corpus tools is expected to greatly streamline corpus analysis.</p>
        <p>Finally, ChatGPT can also be a valuable tool for refining terminological definitions. After
drafting a definition, terminologists can ask ChatGPT to assess and suggest enhancements.
This addresses both content-related issues, such as errors and biases, and formal aspects such
as spelling and stylistic choices (Example 22). Moreover, ChatGPT’s ability to guess the term
from a definition ofers an innovative approach to evaluating the adequacy of a definition.
Should ChatGPT fail to correctly identify the defined term, this may indicate that the definition
is in need of further refinement, a process in which ChatGPT can also provide assistance
(Example 23).</p>
      </sec>
    </sec>
    <sec id="sec-5">
      <title>4. Conclusions</title>
      <p>
        AI has the potential to revolutionize many aspects of human life, and terminology work is no
exception. Up until now, terminological definitions have been the terminographic element
that has most escaped automation [
        <xref ref-type="bibr" rid="ref33">33</xref>
        ], but GenIA-based tools have radically changed the
situation. They are now able to provide terminological definitions that many users might
consider good enough, which means that they would not consult terminological resources. This
shift challenges the very existence of terminological resources as we know them.
      </p>
      <p>While users requiring reliability are likely to still consult terminological resources, the
distinction between human-created and AI-created content will become increasingly blurred.
ChatGPT and the expected emergence of AI tools tailored specifically for terminology work
are set to substantially boost the eficiency of terminologists who must write definitions. This
increase in productivity will not only allow for the prompt update of terminological resources
to reflect knowledge advancement but can also lead to an improvement in quality by facilitating
the creation of flexible definitions. This means that crafting definitions that accurately reflect the
contextual variability of terms will no longer be constrained by the lengthy and labor-intensive
nature of the process.</p>
      <p>Considering these advantages, the integration of AI into terminology work appears inevitable.
As AI technologies grow increasingly sophisticated, the gap in quality between AI-generated
and human-created content is likely to diminish. To remain competitive and relevant against AI
tools such as ChatGPT, terminologists will need to embrace AI and develop the necessary skills
for its eficient integration into their workflow. Just as using computers for terminography
has become a standard practice, we foresee a future where AI-assisted terminography will be
synonymous with terminography itself.</p>
    </sec>
    <sec id="sec-6">
      <title>Acknowledgments</title>
      <p>This research was carried out as part of projects 2020-NP-267503 funded by Quebec’s Society and
Culture Research Fund, 430-2023-0248 funded by the Social Sciences and Humanities Research
Council of Canada, and PID2020-118369GBI00 funded by the Spanish Ministry of Science and
Innovation.</p>
    </sec>
    <sec id="sec-7">
      <title>A. Appendix</title>
      <p>All the answers to the prompts below were obtained with ChatGPT 4 in March 2024.</p>
      <sec id="sec-7-1">
        <title>Example 1: Asking ChatGPT to provide the definition of a term in a specific textual context</title>
        <p>User
What does “work” mean in “For a particle acted upon only by conservative forces, the work
done is path-independent.”?
ChatGPT
In the context of physics, ”work” refers to the transfer of energy to an object via a force causing
the object to move. The work done on the object is calculated as the product of the force
applied to the object, the distance over which the force is applied, and the cosine of the angle
between the force and the direction of movement.</p>
        <p>When it is said that ”for a particle acted upon only by conservative forces, the work done is
path-independent,” it means that the total work done on the particle as it moves from one point
to another is the same, regardless of the path taken between these points. Conservative forces,
such as gravitational and electrostatic forces, have this property because the work done by
these forces depends only on the initial and final positions of the particle, not on the specific
path taken between these positions. Therefore, the total work done by conservative forces
on a particle in moving from one point to another is determined solely by these points and is
independent of the trajectory followed by the particle.</p>
      </sec>
      <sec id="sec-7-2">
        <title>Example 2: Asking ChatGPT to provide the definition of a term in a specific textual context including background information about the user and the text</title>
        <p>User
I am a terminologist, and I am reading an introductory book on statistics for corpus analysis.
Can you explain what “predictive modeling” means in this sentence of the book: “Several
statistical methods can be used for predictive modeling in corpus linguistics”?
ChatGPT
Predictive modeling in the context of corpus linguistics refers to the use of statistical methods
to analyze textual data (corpus) in order to make predictions about future instances of language
use or to generalize about language patterns. This involves building a model that can predict
an outcome based on the features extracted from the text. Here’s how it typically works:
1. Data Collection and Preparation: A large corpus of text is gathered and prepared for
analysis. This could include cleaning the data, annotating it, and structuring it in a way
that can be used for statistical analysis.
2. Feature Extraction: Characteristics (features) of the text are identified and extracted.</p>
        <p>These features might include word frequency, grammatical structures, n-grams (sequences
of ‘n‘ words), syntactic patterns, etc.
3. Model Training: The extracted features are used to train a statistical model. This
involves using a portion of the corpus to teach the model what the outcomes should
be based on the features. Common methods include regression analysis, decision trees,
machine learning algorithms, and more.
4. Testing and Validation: The model is then tested with another portion of the corpus
(not used in the training phase) to check its accuracy and ability to predict outcomes.
This step is crucial to validate that the model can generalize beyond the specific data it
was trained on.
5. Prediction: Once the model is validated, it can be used to make predictions. For example,
it could predict the likelihood of certain language structures or words appearing in similar
texts, or it might be used to infer characteristics of the text’s authors or the contexts in
which the text is likely to occur.</p>
        <p>In corpus linguistics, predictive modeling can be used for various purposes, such as identifying
future language trends, automating text classification, understanding language evolution, or
improving natural language processing systems.</p>
      </sec>
      <sec id="sec-7-3">
        <title>Example 3: Asking ChatGPT to provide the definition of a term in an image</title>
        <p>including background information about the user7
User
What does “weight” mean in this image? Explain in simple terms because I have no background
7Image by Chrislb (Diagram of an artificial neuron). Licensed under the Creative Commons Attribution-Share Alike
3.0 Unported license. URL: https://commons.wikimedia.org/wiki/File:ArtificialNeuronModel_english.png
in artificial intelligence.</p>
        <p>ChatGPT
The image you’ve uploaded shows a diagram of a single neuron within a neural network, which
is a fundamental concept in artificial intelligence. Here’s a simple breakdown of what ”weight”
means in this context:
Think of a neural network like a team of decision-makers, where each ”neuron” is an individual
decision-maker. Each decision-maker (neuron) gets a set of inputs, which are pieces of
information. The ”weights” are like the importance each decision-maker assigns to each piece
of information when making a decision.</p>
        <p>For example, if the decision is about whether to take an umbrella when leaving the house,
one piece of input information might be the weather forecast saying it’s going to rain. If a
particular neuron thinks the weather forecast is very important, it will assign a high weight to
that information. On the other hand, if another input is the fact that it rained yesterday, and the
neuron doesn’t think that’s very important, it might assign a low weight to that information.
So, ”weights” adjust how much influence each piece of input information has on the decision
each neuron makes. After considering all the inputs and their respective weights, the neuron
then decides what output to provide, which contributes to the final decision made by the whole
neural network.</p>
      </sec>
      <sec id="sec-7-4">
        <title>Example 4: Asking ChatGPT to provide the definition of a term in a specific situation</title>
        <p>User
My cousin is a truck driver and said something about a “sideboard”. What does it mean?
ChatGPT
A ”sideboard” in the context of truck driving usually refers to an attachment on the sides
of a truck bed or trailer that helps contain the load and prevent it from spilling or shifting
during transport. These sideboards are often made of wood or metal and can be either fixed or
removable. They efectively increase the height of the sides of the truck bed or trailer, allowing
for more secure transportation of larger or more voluminous cargo.</p>
      </sec>
      <sec id="sec-7-5">
        <title>Example 5: Asking ChatGPT to provide the definition of a term</title>
        <p>User
Define “spectroscopy”.</p>
        <p>ChatGPT
Spectroscopy is a scientific technique used to analyze the properties of light and other forms of
electromagnetic radiation as they interact with matter. It involves measuring the spectrum of
light—that is, the distribution of light intensity with respect to wavelength or frequency—from
an object or sample. By examining these spectra, scientists can infer various characteristics
of the matter being studied, such as its composition, temperature, density, and chemical structure.
Spectroscopy is used in many fields, including chemistry, physics, astronomy, and biology, to
identify substances, study atomic and molecular structures, and understand physical processes.
Diferent types of spectroscopy, such as infrared, ultraviolet, and X-ray spectroscopy, utilize
diferent regions of the electromagnetic spectrum to provide information about the matter under
investigation.</p>
      </sec>
      <sec id="sec-7-6">
        <title>Example 6: Asking ChatGPT to provide the terminological definition of a term</title>
        <p>User
Write a terminological definition of “spectroscopy”.</p>
        <p>ChatGPT
Spectroscopy is a scientific technique used to analyze the interactions between matter and
electromagnetic radiation. It involves measuring the absorption, emission, or scattering of light
by materials to determine their composition, structure, and physical properties. Spectroscopy
can identify chemical elements and compounds, determine concentrations, and reveal molecular
energy levels and mechanisms. This method is fundamental in various fields, including
chemistry, physics, astronomy, and environmental science, ofering insights into the characteristics
of diferent substances and celestial objects.</p>
      </sec>
      <sec id="sec-7-7">
        <title>Example 7: Asking ChatGPT to provide the terminological definition of a term specifying that the defined term should not be mentioned at the beginning of the definition</title>
        <p>User
Write a terminological definition of ”spectroscopy”. Do not mention the defined term at the
beginning of the definition.</p>
        <p>ChatGPT
A scientific technique that involves analyzing the interaction between matter and
electromagnetic radiation to identify, quantify, and study the properties of substances. This method can
be applied across a broad spectrum of wavelengths, including infrared, visible, ultraviolet, and
X-ray regions. It is used to infer the physical properties, chemical composition, and structure of
the analyzed material by examining its emitted, absorbed, or scattered light.</p>
      </sec>
      <sec id="sec-7-8">
        <title>Example 8: Asking ChatGPT to provide the terminological definition of a term specifying contextual and functional constraints</title>
        <p>User
Write a terminological definition of “hydraulic fracturing” intended for high school students
from the thematic perspective of Energy Engineering and the geographic perspective of Canada.
Do not mention the defined term at the beginning of the definition. The definition will be
inserted in a specialized dictionary for high school students.</p>
        <p>ChatGPT
A technique used in the energy industry to extract oil and natural gas from deep underground,
this process involves injecting high-pressure fluid into rock formations. This fluid creates
fractures in the rock, allowing the trapped oil or gas to flow to the surface. In Canada, particularly
in regions like Alberta and British Columbia, it has significantly contributed to the country’s
energy production, tapping into vast reserves that were previously inaccessible. This method
has transformed the energy landscape, making Canada one of the top producers of natural gas
and oil in the world.</p>
      </sec>
      <sec id="sec-7-9">
        <title>Example 9: Asking ChatGPT to provide the terminological definition of a term specifying contextual constraints and allowing for more than one definition if the term is polysemic</title>
        <p>User
Write a terminological definition of “feed” from the thematic perspective of Agriculture. If the
term is polysemic, write one definition per concept associated to the term.</p>
        <p>ChatGPT
In agriculture, feed refers to the food given to domestic animals, particularly livestock, in the
course of animal husbandry. The term is polysemic and has several definitions based on its
diferent uses within this thematic perspective:
1. Animal Nutrition: As a noun, feed encompasses all types of food given to animals,
including hay, grain, and commercially prepared animal feed. This type of feed is specifically
formulated to provide essential nutrients that ensure the health, growth, and productivity
of livestock and other animals.
2. Feeding Process: As a verb, feed means the act of providing food to animals. This
involves the physical process of delivering food to animals, which can vary from manual
distribution to automated feeding systems in a controlled and often scheduled manner to
meet the nutritional needs of animals.
3. Economic Good: In an economic context within agriculture, feed refers to the commodity
that is bought and sold for the purpose of animal consumption. The quality, composition,
and price of feed can significantly impact the agricultural sector, especially in livestock
production.</p>
      </sec>
      <sec id="sec-7-10">
        <title>Example 10: Asking ChatGPT to provide the definition of several terms based on a definitional template</title>
        <p>User
I have the following template to create the definition of concepts belonging to the category
HARD COASTAL DEFENSE STRUCTURE:
type_of:
located_at:
made_of:
has_function:</p>
        <sec id="sec-7-10-1">
          <title>This is the template as applied to the concept GROYNE:</title>
        </sec>
        <sec id="sec-7-10-2">
          <title>Concept: GROYNE</title>
          <p>Definition: hard coastal defense structure made of concrete, wood, steel and/or rock
perpendicular to the shoreline built to protect a shore area, retard littoral drift, reduce longshore
transport, and prevent beach erosion.</p>
        </sec>
        <sec id="sec-7-10-3">
          <title>Template:</title>
          <p>type_of: hard coastal defense structure
located_at: perpendicular to shoreline
made_of: concrete, wood, steel, and/or rock
has_function: shore protection, littoral drift retardation, longshore transport reduction, and
beach erosion prevention
Complete the template of the concepts JETTY, MOLE, and DIKE one by one. And write the
definition of each concept after its template.</p>
          <p>ChatGPT
Let’s start with the concept of a JETTY.</p>
          <p>Concept: JETTY
Template:
• type_of: hard coastal defense structure
• located_at: alongside shorelines or entrances to harbors and ports
• made_of: stone, concrete, and/or wood
• as_function: protection of harbors and coastline, control of sediment transport, and
provision of safe berthing areas
Definition: A jetty is a hard coastal defense structure constructed of stone, concrete, or wood,
extending from the shore into the water, often at harbor entrances or river mouths. It is designed
to protect the shoreline and harbor areas by controlling sediment transport, preventing shoreline
erosion, and providing safe mooring areas.</p>
          <p>Next, we’ll cover the concept of a MOLE.</p>
          <p>Concept: MOLE
Template:
• type_of: hard coastal defense structure
• located_at: extending into the sea from a coast or harbor
• made_of: large stones, concrete, and/or wood
• has_function: harbor protection, wave break, and creation of calm waters for docks and
marinas
Definition: A mole is a massive hard coastal defense structure made of large stones, concrete,
or wood, extending from the coast or harbor into the sea. Its primary functions are to protect
harbors, act as a breakwater to reduce the force of waves, and create calm waters for docks,
marinas, and anchorages.</p>
          <p>Finally, let’s define a DIKE.</p>
          <p>Concept: DIKE
Template:
• type_of: hard coastal defense structure
• located_at: along coastlines or riverbanks
• made_of: earth, clay, concrete, and/or rock
• has_function: flood prevention, erosion control, and land reclamation
Definition: A dike is a hard coastal defense structure built along coastlines or riverbanks,
constructed from materials such as earth, clay, concrete, or rock. It is designed to prevent flooding,
control erosion, and reclaim land from bodies of water, thereby protecting and maintaining the
integrity of inland areas.</p>
        </sec>
      </sec>
      <sec id="sec-7-11">
        <title>Example 11: Asking ChatGPT to explain the diference between two terms</title>
        <p>User
What is the diference between “quantum decoherence” and “wave function collapse”?
ChatGPT
Quantum decoherence and wave function collapse are related concepts in quantum mechanics,
but they refer to diferent processes:
• Quantum Decoherence: This is the process by which a quantum system loses its
quantum properties and behavior, transitioning towards classical behavior due to the
interaction with its environment. Decoherence explains how the superposition of states
(where a system exists in multiple states simultaneously) appears to reduce to a single
outcome. However, decoherence does not predict which outcome is realized; it only
describes how coherent superpositions are lost, leaving a mixture of states rather than a
single definite state. Essentially, decoherence provides a mechanism for the classical-like
behavior of macroscopic systems by showing how they lose their quantum coherence
through interactions with their surroundings, leading to the appearance of wave function
collapse without actually collapsing the wave function.
• Wave Function Collapse: This concept is used to describe the change of a quantum
system from a superposition of states to a single eigenstate due to a measurement or
observation. Wave function collapse is a fundamental concept in the Copenhagen
interpretation of quantum mechanics, where the act of measurement forces a quantum system
to settle into one of the possible eigenstates. The wave function collapse is instantaneous
and discontinuous, and it selects one outcome from the probabilities described by the
wave function.</p>
        <p>In summary, quantum decoherence describes how interactions with the environment cause a
quantum system to appear classical, leading to a loss of quantum superposition, but without
specifying which state the system will end up in. On the other hand, wave function collapse
refers to the process where a quantum system suddenly transitions from a superposition of
states to a single definite state upon measurement. While decoherence can be seen as a part of
the process leading to collapse by efectively ”hiding” the other outcomes from the superposition,
it does not by itself explain why a specific outcome is observed in a measurement.</p>
      </sec>
      <sec id="sec-7-12">
        <title>Example 12: Asking for the explanation of a sentence</title>
        <p>User
Explain this sentence in simple terms: “Energy levels in quantum systems are generally the real
eigenvalues of their Hermitian Hamiltonian”.</p>
        <p>ChatGPT
This sentence is about how energy is described in quantum mechanics, which is the branch of
physics dealing with very small particles like atoms and electrons.</p>
        <p>• Quantum systems: These are systems studied in quantum mechanics, like atoms or
particles, where the rules of classical physics don’t apply.
• Energy levels: In quantum systems, energy doesn’t vary smoothly but comes in discrete
steps or ”levels.” Just like you can only climb stairs one step at a time, particles in quantum
systems can only have specific energy levels.
• Real eigenvalues: This term comes from mathematics, particularly from the study
of matrices. Eigenvalues are numbers that give important information about a system.
In this context, ”real” means these numbers are not complex (i.e., they don’t involve
imaginary numbers, which are numbers involving the square root of negative one).
• Hermitian Hamiltonian: This is a fancy way of saying a specific type of mathematical
function or operator that is used in quantum mechanics to describe the total energy of a
system. ”Hermitian” means it has certain symmetry properties, and ”Hamiltonian” refers
to the total energy of the system.</p>
        <p>So, putting it all together, the sentence means that in quantum mechanics, the specific energy
values that a quantum system can have are determined by solving a mathematical problem
involving the system’s Hamiltonian, which is a function that describes its total energy. The
solutions to this problem are the energy levels, and they are real numbers, not complex numbers.</p>
      </sec>
      <sec id="sec-7-13">
        <title>Example 13: Asking ChatGPT to provide a list of all the semantic features in a set of definitions of the same term</title>
        <p>User
Make a list of all the semantic features of ”radiative forcing” contained in these definitions.
Each semantic feature should be a sentence. Indicate in brackets the source that contains each
semantic trait. Be exhaustive: include all possible traits, even those mentioned in only one
definition, but if a trait is mentioned in more than one definition, list all sources.
A change in average net radiation at the top of the troposphere (known as the tropopause)
because of a change in either incoming solar or exiting infrared radiation. A positive radiative
forcing tends on average to warm the earth’s surface; a negative radiative forcing on average
tends to cool the earth’s surface. Greenhouse gases, when emitted into the atmosphere, trap
infrared energy radiated from the earth’s surface and therefore tend to produce positive
radiative forcing. (EIA: https://www.eia.gov/tools/glossary/)
A change in the balance between incoming solar radiation and outgoing infrared (i.e., thermal)
radiation. Without any radiative forcing, solar radiation coming to the Earth would continue to
be approximately equal to the infrared radiation emitted from the Earth. The addition of
greenhouse gases to the atmosphere traps an increased fraction of the infrared radiation, reradiating
it back toward the surface of the Earth and thereby creates a warming influence. (UNFCCC:
https://unfccc.int/resource/cd_roms/na1/ghg_inventories/english/8_glossary/Glossary.htm)
Change in the balance between incoming solar radiation and outgoing infrared
radiation. Causes include internal changes and external forcing, such as changes in solar
output or carbon dioxide concentrations. Without any radiative forcing, solar
radiation coming to the Earth would approximately equal to the infrared radiation emitted
from Earth. A positive forcing warms the Earth, a negative forcing cools it. (CARA:
https://web.archive.org/web/20060901114821/http://www.cara.psu.edu/tools/glossary.asp)
A change in the balance between incoming solar radiation and outgoing infra-red radiation.
Without any radiative forcing, solar radiation coming to the Earth would continue to be
approximately equal to the infra-red radiation emitted from the Earth. The addition of
greenhouse gases traps and increased fraction of the infra-red radiation, reradiating it
back toward the surface and creating a warming influence (i.e., positive radiative forcing
because incoming solar radiation will exceed outgoing infra-red radiation). (EO-NASA:
https://earthobservatory.nasa.gov/glossary/q/s)
A change in average net radiation (in W m-2) at the top of the troposphere resulting from a
change in either solar or infrared radiation due to a change in atmospheric greenhouse gases
concentrations; perturbance in the balance between incoming solar radiation and outgoing
infrared radiation. (IPCC: https://archive.ipcc.ch/ipccreports/sres/aviation/index.php?idp=158)
The term radiative forcing refers to changes in the energy balance of the earth-atmosphere
system in response to a change in factors such as greenhouse gases, land-use change, or
solar radiation. The climate system inherently attempts to balance incoming (e.g., light) and
outgoing (e.g. heat) radiation. Positive radiative forcings increase the temperature of the lower
atmosphere, which in turn increases temperatures at the Earth’s surface. Negative radiative
forcings cool the lower atmosphere. Radiative forcing is most commonly measured in units of
watts per square meter (W/m2). (Climate CoLab: https://www.climatecolab.org/wiki/Glossary)
A change imposed upon the climate system which modifies the radiative balance of that
system. The causes of such a change may include changes in the sun, clouds, ice, greenhouse
gases, volcanic activity, and other agents. Radiative forcing is often specified as the net
change in energy flux at the troposphere (watts per square meter). Radiative forcing may
sometimes be referred to as external forcing or perturbations of the climate. (GCMD-NASA:
https://gcmd.earthdata.nasa.gov/KeywordViewer/)
The net change in the energy balance of the Earth system due to an external perturbation,
measured in units of watts per square metre (W/m2). (CCCR: https://natural-resources.canada.
ca/sites/www.nrcan.gc.ca/files/energy/Climate-change/pdf/CCCR_Definitions-EN-111919.pdf)
Diference between incoming solar radiation on the Earth and outgoing thermal radiation from
the Earth. (ISO 22948:2020: https://www.iso.org/obp/ui#iso:std:iso:22948:ed-1:v1:en:term:3.5.4)
A measure of the influence of a particular factor (e.g. greenhouse gas (GHG),
aerosol, or land use change) on the net change in the Earths energy balance. (EPA:
https://19january2017snapshot.epa.gov/climatechange/glossary-climate-change-terms_.html)
A disturbance to the earth’s energy balance. (Termium Plus: https://www.btb.termiumplus.gc.
ca/)
A perturbation of the global radiation balance. This may be human-induced or natural.
(LennTech: https://www.lenntech.com/greenhouse-effect/climate-change-glossary.htm)
A change in the balance between incoming solar radiation and outgoing infrared radiation.
Without any radiative forcing, solar radiation coming to the Earth would continue to
be approximately equal to the infrared radiation emitted from the Earth. The addition
of greenhouse gases traps an increased fraction of the infrared radiation, radiating it
back toward the surface and creating a warming influence (i.e., positive radiative forcing
because incoming solar radiation will exceed outgoing infrared radiation). (NAL: https:
//agclass.nal.usda.gov/vocabularies/nalt/concept?uri=https://lod.nal.usda.gov/nalt/281696)
Changes in energy fluxes F (in W/m2) at the top-of-the-atmosphere that cause climate change.
It is defined as positive (negative) if it leads to warming (cooling). The radiative forcing for
a doubling of CO2 is ΔF2xCO2 = 3.7 W/m2. Other examples are increased solar radiation
(positive), increased aerosols (negative) or increased surface albedo (negative), e.g. due to land
use changes. (OSU: https://open.oregonstate.education/climatechange/back-matter/glossary/)
The efect of greenhouse gases on the radiation of heat in the atmosphere. (Wiktionary:
https://en.wiktionary.org/)
Radiative forcing is the change in energy flux in the atmosphere caused by natural or
anthropogenic factors of climate change as measured in watts per meter squared. (Wikipedia:
https://en.wikipedia.org/)
The diference between sunlight absorbed by Earth and energy radiated back to space,
disrupted by greenhouse gas levels. (Footprint Consulting: https://footprintconsulting.org/
climate-knowledge-base/)
ChatGPT
• Radiative forcing occurs at the top of the troposphere, specifically at the tropopause. (EIA,</p>
        <p>IPCC)
• It involves a change in average net radiation due to variations in incoming solar or
outgoing infrared radiation. (EIA, IPCC)
• Positive radiative forcing tends to warm the Earth’s surface, while negative radiative
forcing tends to cool it. (EIA, CARA, OSU)
• Greenhouse gases trap infrared energy radiated from the Earth’s surface, leading to
positive radiative forcing. (EIA, UNFCCC, EO-NASA)
• Radiative forcing afects the balance between incoming solar radiation and outgoing
infrared radiation. (UNFCCC, CARA, EO-NASA, IPCC, NAL)
• Without radiative forcing, solar radiation coming to Earth would be approximately equal
to the infrared radiation emitted from Earth. (UNFCCC, CARA, EO-NASA, NAL)
• Causes of radiative forcing include internal changes and external factors like solar output
or carbon dioxide concentrations. (CARA)
• Greenhouse gases increase the fraction of trapped infrared radiation, reradiating it back
toward the Earth’s surface, creating a warming efect. (UNFCCC, EO-NASA, NAL)
• It is measured in watts per square meter (W/m2̂). (IPCC, Climate CoLab, GCMD-NASA,</p>
        <p>CCCR, OSU)
• Radiative forcing reflects changes in the energy balance of the earth-atmosphere system
due to various factors such as greenhouse gases and solar radiation. (Climate CoLab,
GCMD-NASA)
• The climate system inherently tries to balance incoming and outgoing radiation. (Climate</p>
        <p>CoLab)
• It can increase or decrease the temperature of the lower atmosphere, afecting the Earth’s
surface temperature. (Climate CoLab)
• Radiative forcing can be imposed by changes in the sun, clouds, ice, greenhouse gases,
volcanic activity, and other agents. (GCMD-NASA)
• It is often specified as the net change in energy flux at the troposphere. (GCMD-NASA)
• It may sometimes be referred to as external forcing or perturbations of the climate.</p>
        <p>(GCMD-NASA)
• Represents the net change in the Earth system’s energy balance due to external
perturbation. (CCCR)
• Described as the diference between incoming solar radiation on the Earth and outgoing
thermal radiation from the Earth. (ISO 22948:2020)
• Indicates the influence of a particular factor on the net change in Earth’s energy balance.</p>
        <p>(EPA)
• Considered a disturbance to the Earth’s energy balance. (Termium Plus)
• Perturbs the global radiation balance, either human-induced or natural. (LennTech)
• Changes in energy fluxes at the top-of-the-atmosphere cause climate change, defined as
positive for warming and negative for cooling. (OSU)
• Relates to the efect of greenhouse gases on the radiation of heat in the atmosphere.</p>
        <p>(Wiktionary)
• It is the change in energy flux in the atmosphere caused by natural or anthropogenic
factors. (Wikipedia)
• Involves the disruption of the balance between absorbed sunlight by Earth and energy
radiated back to space due to greenhouse gas levels. (Footprint Consulting)</p>
      </sec>
      <sec id="sec-7-14">
        <title>Example 14: Continuation of the conversation in Example 13. Asking ChatGPT to verify the accuracy of each semantic trait</title>
        <p>User
Verify exhaustively the accuracy of each semantic trait. Copy each semantic trait, add the
source(s), and then label it as accurate or inaccurate according to your knowledge. If the
semantic trait is accurate, do not write anything else. If the semantic trait is inaccurate, write
an explanation and propose a corrected version of the trait.</p>
        <p>ChatGPT
12. It can increase or decrease the temperature of the lower atmosphere, afecting
the Earth’s surface temperature. (Climate CoLab)</p>
        <p>Accurate
13. Radiative forcing can be imposed by changes in the sun, clouds, ice, greenhouse
gases, volcanic activity, and other agents. (GCMD-NASA)</p>
        <p>Accurate
14. It is often specified as the net change in energy flux at the troposphere.
(GCMDNASA)
This trait is slightly inaccurate because radiative forcing is commonly measured at the
top of the atmosphere, not just at the troposphere.</p>
        <p>Corrected version: Radiative forcing is often specified as the net change in energy flux at
the top of the atmosphere.
15. It may sometimes be referred to as external forcing or perturbations of the
climate. (GCMD-NASA)</p>
        <p>Accurate
16. Represents the net change in the Earth system’s energy balance due to external
perturbation. (CCCR)</p>
        <p>Accurate
17. Described as the diference between incoming solar radiation on the Earth and
outgoing thermal radiation from the Earth. (ISO 22948:2020)</p>
        <p>Accurate
18. Indicates the influence of a particular factor on the net change in Earth’s energy
balance. (EPA)</p>
        <p>Accurate
19. Considered a disturbance to the Earth’s energy balance. (Termium Plus)</p>
        <p>Accurate
20. Perturbs the global radiation balance, either human-induced or natural.
(LennTech)</p>
        <p>Accurate
21. Changes in energy fluxes at the top-of-the-atmosphere cause climate change,
defined as positive for warming and negative for cooling. (OSU)</p>
        <p>Accurate
22. Relates to the efect of greenhouse gases on the radiation of heat in the
atmosphere. (Wiktionary)</p>
        <p>Accurate
23. It is the change in energy flux in the atmosphere caused by natural or
anthropogenic factors. (Wikipedia)
This trait is somewhat imprecise, as it doesn’t clearly state that the change occurs at the
top of the atmosphere.</p>
        <p>Corrected version: Radiative forcing is the change in energy flux at the top of the
atmosphere caused by natural or anthropogenic factors.
24. Involves the disruption of the balance between absorbed sunlight by Earth and
energy radiated back to space due to greenhouse gas levels. (Footprint Consulting)
Accurate</p>
      </sec>
      <sec id="sec-7-15">
        <title>Example 15: Continuation of the conversation in Example 14. Asking ChatGPT to classify the semantic traits based on relevance</title>
        <p>User
Classify all of the traits in three categories (highly relevant, relevant, less relevant) according to
the relevance to define ”radiative forcing” in a terminological resource about climate change. If
a trait has been corrected, use the corrected version.</p>
        <p>ChatGPT
Highly Relevant:
Relevant:
• Radiative forcing occurs at the top of the troposphere, specifically at the tropopause. (EIA,</p>
        <p>IPCC)
• It involves a change in average net radiation due to variations in incoming solar or
outgoing infrared radiation. (EIA, IPCC)
• Positive radiative forcing tends to warm the Earth’s surface, while negative radiative
forcing tends to cool it. (EIA, CARA, OSU)
• Radiative forcing afects the balance between incoming solar radiation and outgoing
infrared radiation. (UNFCCC, CARA, EO-NASA, IPCC, NAL)
• Greenhouse gases trap infrared energy radiated from the Earth’s surface, leading to
positive radiative forcing. (EIA, UNFCCC, EO-NASA)
• Without radiative forcing, solar radiation coming to Earth would be approximately equal
to the infrared radiation emitted from Earth. (UNFCCC, CARA, EO-NASA, NAL)
• It is measured in watts per square meter (W/m2̂). (IPCC, Climate CoLab, GCMD-NASA,</p>
        <p>CCCR, OSU)
• Represents the net change in the Earth system’s energy balance due to external
perturbation. (CCCR)
• Causes of radiative forcing include internal changes and external factors like solar output
or carbon dioxide concentrations. (CARA)
• Greenhouse gases increase the fraction of trapped infrared radiation, reradiating it back
toward the Earth’s surface, creating a warming efect. (UNFCCC, EO-NASA, NAL)
• The climate system inherently tries to balance incoming and outgoing radiation. (Climate</p>
        <p>CoLab)
• It can increase or decrease the temperature of the lower atmosphere, afecting the Earth’s
surface temperature. (Climate CoLab)
• Radiative forcing can be imposed by changes in the sun, clouds, ice, greenhouse gases,
volcanic activity, and other agents. (GCMD-NASA)
• It may sometimes be referred to as external forcing or perturbations of the climate.</p>
        <p>(GCMD-NASA)
• It is often specified as the net change in energy flux at the troposphere. (GCMD-NASA)
Less Relevant:
• Indicates the influence of a particular factor on the net change in Earth’s energy balance.</p>
        <p>(EPA)
• Considered a disturbance to the Earth’s energy balance. (Termium Plus)
• Perturbs the global radiation balance, either human-induced or natural. (LennTech)
• Changes in energy fluxes at the top-of-the-atmosphere cause climate change, defined as
positive for warming and negative for cooling. (OSU)
• Relates to the efect of greenhouse gases on the radiation of heat in the atmosphere.</p>
        <p>(Wiktionary)
• It is the change in energy flux in the atmosphere caused by natural or anthropogenic
factors. (Wikipedia)
• Involves the disruption of the balance between absorbed sunlight by Earth and energy
radiated back to space due to greenhouse gas levels. (Footprint Consulting)
• Described as the diference between incoming solar radiation on the Earth and outgoing
thermal radiation from the Earth. (ISO 22948:2020)</p>
      </sec>
      <sec id="sec-7-16">
        <title>Example 16: Continuation of the conversation in Example 15. Asking ChatGPT to propose semantic traits that should be included in the list</title>
        <p>User
Are there any semantic traits that should be included that are currently absent from the list?
ChatGPT
To comprehensively define ”radiative forcing” in the context of climate change, the list could be
enhanced by including the following semantic traits that are currently absent or not explicitly
mentioned:
• Quantification of the efect of radiative forcing on climate systems: While it’s
stated that radiative forcing is measured in watts per square meter, the direct impact on
global temperature or climate change over time could be more explicitly addressed. This
includes how specific quantities of radiative forcing correlate with temperature changes
on Earth.
• Time scale of the impact of radiative forcing: The list does not explicitly mention
how the efects of radiative forcing might vary over diferent time scales, which is crucial
for understanding its role in climate change dynamics.
• Comparative scale of diferent forcings: An indication of the relative strength or
importance of diferent types of radiative forcing (e.g., comparing the forcing from CO2
emissions to that from solar irradiance changes) would provide more context on the
various sources and their impacts.
• Role of feedback mechanisms: The interaction between radiative forcing and climate
feedback mechanisms (like water vapor feedback, ice-albedo feedback) is crucial for a full
understanding of the climate system’s response to radiative forcing.
• Distinction between direct and indirect radiative forcing: Clarifying the diference
between direct radiative forcing (e.g., from CO2, solar irradiance) and indirect radiative
forcing (e.g., changes in cloud cover, aerosol interactions) would provide a more nuanced
understanding of the concept.</p>
        <p>Adding these traits would give a more complete and nuanced understanding of radiative forcing
and its significance in the study of climate change.</p>
      </sec>
      <sec id="sec-7-17">
        <title>Example 17: Continuation of the conversation in Example 16. Asking ChatGPT to write a new definition based on the list of semantic traits</title>
        <p>User
Write a new definition of ”radiative forcing” for a terminological resource about climate change
based on the list of semantic traits. Justify your definition and list the sources.
ChatGPT
Radiative forcing is a measure of the change in the energy balance of the Earth’s climate system,
quantified as the diference in watts per square meter (W/m2̂) between incoming solar radiation
and outgoing infrared radiation at the top of the troposphere, particularly at the tropopause.
It is a key factor in climate dynamics, determining the Earth’s surface temperature by either
warming or cooling the planet. Positive radiative forcing, often resulting from increased
concentrations of greenhouse gases, enhances the greenhouse efect by trapping more infrared
radiation emitted from the Earth, leading to a net increase in absorbed energy and a warming
efect. Conversely, negative radiative forcing, which can be caused by factors like increased
aerosol concentrations or changes in surface albedo, results in a net loss of energy and a cooling
efect. The magnitude of radiative forcing is central to predicting the rate and extent of climate
change, as it influences both the immediate energy balance and long-term climate feedback
mechanisms.</p>
        <p>Justification:
• Energy balance focus: The definition emphasizes the concept of energy balance, which
is central to understanding radiative forcing and its climatic implications.
• Quantification and location: Specifying the measurement in watts per square meter
at the top of the troposphere provides a clear, quantifiable description that aligns with
scientific standards.
• Cause and efect relationship: It addresses the cause (greenhouse gases, aerosols,
albedo changes) and efect (warming or cooling) directly, highlighting the role of radiative
forcing in climate change.
• Feedback mechanisms consideration: Although not explicitly detailed in the traits
list, the mention of climate feedback mechanisms acknowledges the complex interactions
within the climate system that radiative forcing afects.
• EIA, IPCC: for the concept of radiative forcing occurring at the top of the troposphere
and its quantification.
• UNFCCC, CARA, EO-NASA, IPCC, NAL: for the explanation of how radiative forcing
alters the balance between incoming and outgoing radiation.
• Climate CoLab, GCMD-NASA, CCCR: for the broader context of radiative forcing’s role
in the climate system and its measurement units.
• OSU: for the distinction between positive and negative radiative forcing and its impact
on global climate change.</p>
      </sec>
      <sec id="sec-7-18">
        <title>Example 18: Asking ChatGPT to analyze some concordance lines</title>
        <p>User
Analyze the following concordance lines of the term ”biological phosphorus removal” from a
semantic point of view. What semantic traits of the term can you extract from them? Always
mention the lines to justify your answer.</p>
        <p>
          1. The parent sequencing batch reactor was operated over 100 d with synthetic wastewater
(see Supporting Information for details) and achieved stable biological phosphorus removal
eficiency (.99%).
2. By contrast, metagenomics of reactors with certain intentionally enhanced functions, for
example, enhanced biological phosphorus removal reactor (EBPR), cellulose-degrading
reactor, phenol decomposing reactor, sludge digester etc., makes more practical sense for
most research institutions lack of such admirable resources, and thus is crucial for wide
application of metagenomic techniques.
3. Key words: Bioreactor, biofilm carriers, nitrogen removal, biological phosphorus removal,
sewage, treatment reliable for organic carbon and nutrients removal without some of the
problems of activated sludge processes[
          <xref ref-type="bibr" rid="ref38 ref5">5</xref>
          ].
4. Fig. 3: The average efluent soluble COD concentration from each reactor during
experiment Phosphorus removal: Biological P-removal using enhanced biological phosphorus
removal (EBPR) was carried out in this study.
5. Phosphorus can also be removed through enhanced biological phosphorus removal (EBPR),
however, the successful operation of EBPR depends on many process operational factors,
especially variations in wastewater quality [
          <xref ref-type="bibr" rid="ref14">14</xref>
          ].
6. Biological phosphorus removal (BPR) [
          <xref ref-type="bibr" rid="ref2 ref35">2</xref>
          ], chemical precipitation, and adsorption by
functionalized materials [
          <xref ref-type="bibr" rid="ref3 ref36">3</xref>
          ] are commonly used for phosphorus removal.
7. In Concept 1, additional mixing energy of 5 MJ/kg Premoved originated from the biological
phosphorus removal [
          <xref ref-type="bibr" rid="ref40 ref7">7</xref>
          ].
8. Another practical aspect worth highlighting in this section is the use of an enhanced
biological phosphorus removal (EBPR) for phosphorus removal from wastewaters.
9. ” (12) ”Degradation of diuron by Phanerochaete chrysosporium: role of ligninolytic
enzymes and cytochrome P450” by J. da Silva Coelho-Moreira et al. (13) ”Dynamics
of intracellular polymers in enhanced biological phosphorus removal processes under
diferent organic carbon concentrations” by L.
10. Biological Treatment of Wastewater by Sequencing Batch Reactors such as biological
phosphorus removal, aerobic oxidation of nitrogen (nitrification) and anoxic elimination
of nitrate (denitrification).
11. In the biological phosphorus removal, the phosphorus in the influent wastewater is
incorporated into cell biomass, which subsequently is removed from the process as a
result of sludge wasting.
12. Our approach included the creation and use of domain specific HMMs to find members of
the type I PKS domain in six published metagenomic data sets - Minnesota farm soil (MSF)
[
          <xref ref-type="bibr" rid="ref21">21</xref>
          ], Sargasso Sea (SGS) [
          <xref ref-type="bibr" rid="ref22">22</xref>
          ], human gut (HGUT) [
          <xref ref-type="bibr" rid="ref23">23</xref>
          ], acid mine drainage (AMD) [
          <xref ref-type="bibr" rid="ref24">24</xref>
          ],
enhanced biological phosphorus removal sludges (EBPRS) [
          <xref ref-type="bibr" rid="ref25">25</xref>
          ] and whale falls (bones
from sunken whales) (WLF) [
          <xref ref-type="bibr" rid="ref21">21</xref>
          ].
13. In nearly all seven data sets the KS domain is found most frequently (with the exception
of enhanced biological phosphorus removal sludge data sets) followed by the AT, PP or
KR domains.
14. A2O-activated sludge systems were composed of an anaerobic/anoxic/aerobic process
with simultaneous nitrification-denitrification and biological phosphorus removal .
15. Biological phosphorus removal in wastewater treatment systems using the enhanced
biological phosphorus removal (EBPR) process is increasingly being used as an alternative
to chemical precipitation processes [
          <xref ref-type="bibr" rid="ref41 ref8">8</xref>
          ].
16. Biological phosphorus removal in wastewater treatment systems using the enhanced
biological phosphorus removal (EBPR) process is increasingly being used as an alternative
to chemical precipitation processes [
          <xref ref-type="bibr" rid="ref41 ref8">8</xref>
          ].
17. Both plants are configured for biological phosphorus removal and have stable
performance.
18. Determination was made in the first stage that Lepothrix sppwas present at 30 °C, while
20° acetate was consumed in anaerobic condition and biological phosphorus removal
by Rhodocyclacae-afiliating OUT 14 at 30 and 35° Celsisus acetate was completed and
phosphorus removal did not occur as high [71].
19. WWTPs can be engineered to enable enhanced biological phosphorus removal (EBPR).
20. Engineered systems performing enhanced biological phosphorus removal (EBPR), ofer
stable, lower complexity communities for studying phage-host interactions.
21. One such system, wastewater treatment, relies on a process known as enhanced biological
phosphorus removal (EBPR; [
          <xref ref-type="bibr" rid="ref37 ref4">4</xref>
          ]) to remove dissolved organic carbon and phosphorus.
22. Gram-positive Tetrasphaera-related organisms (Actinobacteria) are putative
polyphosphate-accumulating organisms that are abundant in many fullscale
enhanced biological phosphorus removal plants [55] and Haliscomenobacter are one of the
iflamentous bacteria detected in wastewater treatment plants which are responsible for
biomass bulking [56,57].
23. Nowadays, enhanced biological phosphorus removal (EBPR) has been broadly applied
in wastewater containing phosphorus (P) treatment because of its lower cost and more
environmental friendly technology compared to chemical precipitation or adsorption [
          <xref ref-type="bibr" rid="ref1 ref34">1</xref>
          ].
24. The P content of excess sludge in this study, namely, anoxic sludge, was 12.3% of dry
weight significantly higher than that in traditional biological phosphorus removal process
(3%) [
          <xref ref-type="bibr" rid="ref18">18</xref>
          ], which agrees well with the results of chemical analysis, further demonstrating
the existence of phosphorus removal microorganisms in this two-sludge process studied
here. 3.2.
25. Wilmes et al. conducted a series of metaproteomic studies on the molecular mechanisms
of enhanced biological phosphorus removal (EBPR) [
          <xref ref-type="bibr" rid="ref21 ref22 ref37 ref4">4,21,22</xref>
          ].
        </p>
        <p>ChatGPT
From the provided concordance lines, we can extract several semantic traits of ”biological
phosphorus removal” (BPR):
1. Process Eficiency and Stability : Line 1 mentions a reactor achieving ”stable biological
phosphorus removal eficiency,” indicating the process’s reliability and efectiveness.
2. Comparative Advantage: Lines 2 and 23 highlight the practicality and
environmental preference of enhanced biological phosphorus removal (EBPR) over other methods,
indicating a comparative advantage in terms of resource usage and environmental impact.
3. Application in Wastewater Treatment: Many lines (e.g., 3, 10, 11, 15, 17, 19, 21)
associate biological phosphorus removal with wastewater treatment, revealing its primary
application in removing phosphorus from wastewater to prevent pollution.
4. Integration with Other Processes: Lines like 3, 14, and 20 mention the integration of
biological phosphorus removal with other processes like nitrogen removal and
nitrificationdenitrification, suggesting its compatibility and combinatory use in comprehensive waste
treatment systems.
5. Dependency on Operational Factors: Line 5 and 18 discuss how the success of EBPR
depends on operational factors and conditions, such as wastewater quality and temperature,
indicating a sensitivity to environmental and operational variables.
6. Microbial Activity: Lines like 11, 22, and 24 refer to the role of microorganisms in
phosphorus removal, highlighting the biological aspect of the process where specific
bacteria or microbial communities are essential for phosphorus uptake and accumulation.
7. Technological and Research Interest: Lines 2, 9, 12, and 25 show the interest in
researching and enhancing the understanding of the biological mechanisms and eficiency
of phosphorus removal, indicating ongoing scientific and technological development in
this field.
8. Economic and Environmental Benefits : Line 23 specifically mentions the lower cost
and environmental friendliness of EBPR, portraying it as a sustainable and cost-efective
solution for phosphorus removal in wastewater treatment.</p>
        <p>These lines collectively suggest that biological phosphorus removal, particularly in its
enhanced form (EBPR), is a crucial, eficient, and environmentally preferred method for treating
wastewater, with ongoing research and development focused on optimizing its application and
understanding the underlying biological processes.</p>
      </sec>
      <sec id="sec-7-19">
        <title>Example 19: Asking ChatGPT to analyze some concordance lines where the target term is the object of the sentence</title>
        <p>User
Analyze the following concordance lines of the term ”hash” from a semantic point of view.
What semantic traits of the term can you extract from them? Always mention the lines to
justify your answer.
20. Beginning with the public key, this string of values is first ran through the SHA-256
hashing algorithm to produce a hash, and then that hash is computed using RIPEMD160
to produce a bitcoin address.
21. However, if you input the same data again, it will always produce the same hash .
22. Since the hash is generated before the signature is computed, we need to exclude the
signature block of a certificate when computing the hash.
23. Whenever you make a payment in cryptocurrency, you receive a hash displayed in your
wallet.
24. This means the hash is often denoted as a combination of numbers (0 to 9) and letters (a
to f).</p>
        <p>25. Changing the hash stored in block 3 changes the overall data stored inside it.
ChatGPT
Analyzing these lines, several semantic traits of the term ”hash” emerge, which relate to its role
in technology and cryptography:
1. Cryptocurrency and Blockchain Context: Lines 1, 3, 8, 9, 17, 19, and 23 associate
”hash” with the process of cryptocurrency transactions and blockchain technology. The
hash functions as a unique identifier or verification tool in these systems.
2. Data Integrity and Verification: Lines 2, 5, 13, 14, 15, and 16 suggest that a hash is used
to ensure data integrity and verification. For example, line 13 mentions verifying the hash
of each block to ensure data has not been altered, highlighting its role in maintaining the
accuracy and consistency of data.
3. Unique and Consistent Output: Lines 21 and 22 imply that a hash function
produces a unique and consistent output for the same input data, which is fundamental in
cryptographic processes and data verification.
4. Speed and Eficiency: Line 4 mentions that signing a hash can be much faster than
signing long messages, indicating the eficiency and speed that hashing brings to
cryptographic operations.
5. Compact Representation: Lines 6, 10, and 20 discuss how hashes can shorten the size
of data or addresses, suggesting that a hash serves as a compact representation of larger
sets of data.
6. Mining Process: Lines 7 and 19 refer to the process of mining, where a specific hash
value is sought, demonstrating the hash’s role in the computational efort to validate
blockchain transactions.
7. Digital Signatures and Authentication: Lines 6, 9, and 22 highlight the use of hashes
in creating and verifying digital signatures, which are crucial for authentication in
cryptographic protocols.
8. Alphanumeric Representation: Lines 12 and 24 indicate that hashes are often
represented as a combination of numbers and letters, showing their format as alphanumeric
strings.
9. Technical Implementation: Lines 18 and 20 mention specific hashing algorithms like
SHA-256 and RIPEMD160, pointing to the technical aspects of how hashes are generated
and their role in cryptographic functions.</p>
        <p>These lines collectively portray ”hash” as a fundamental concept in cryptography and
digital transactions, emphasizing its importance in ensuring data integrity, facilitating
eficient verification, and enabling secure and authenticated communication.</p>
      </sec>
      <sec id="sec-7-20">
        <title>Example 20: Asking ChatGPT to analyze some concordance lines where the target term holds a causal relation with other terms</title>
        <p>User
Analyze the following concordance lines of the term ”erosion” from a semantic point of view.
What semantic traits of the term can you extract from them? Always mention the lines to
justify your answer.
17. Sediments result from soil erosion in water catchments and are transported to rivers by
water run-of.
18. Erosion problems are often located at urbanized areas because erosion causes the largest
damage in these areas.
19. Section C is characterized by coastal erosion which caused settlement dislocation.
20. Fig. 12 shows how coastal erosion leads to building damage and loss.
21. Erosion causes sedimentation in lakes, reservoirs, and streams, which eventually require
dredging.
22. This erosion will cause higher waves at the toe of the new seawall, with consequently
higher overtopping discharges.
23. This assumption is likely reasonable when the local slope change induced by
erosion/deposition of sediments is small compared to the mean beachface slope.
24. Erosion generally leads to either great monetary losses due to storm damage or even
larger expenditures for shore protection to prevent the loss.
25. This book examines the problems caused by soil erosion from farms, which is the main
cause of nonpoint source pollution.</p>
        <p>ChatGPT
Analyzing these concordance lines of the term ”erosion,” we can extract several semantic traits:
1. Destructive Nature: Erosion is frequently associated with damage and loss. Lines 1, 15,
19, and 20 mention economic loss, ecological damage, societal problems, and settlement
dislocation, emphasizing the destructive impact of erosion on both the environment and
human infrastructure.
2. Geological Process: It involves the wearing away of earth or rock, particularly in the
context of coastal, soil, and tectonic activities. Lines 3, 4, and 6 speak of geological
processes like the formation of flat surfaces, mountain formation, and sediment coverage
on abyssal plains.
3. Human-Induced Factors: Erosion is often exacerbated by human activities such as
agriculture, construction, and industrial operations. Lines 5, 16, and 25 mention soil
erosion from agriculture, construction permits in risk areas, and nonpoint source pollution
from farms.
4. Need for Intervention: There is a recurring theme of the need for human intervention
to manage or mitigate erosion efects, such as artificial nourishment, dune restoration,
and shore protection mentioned in lines 2, 10, and 24.
5. Resultant Features and Processes: Erosion leads to specific geological and
environmental features or processes, such as the creation of peneplains, altiplanos, and sedimentation
in water bodies, as well as the unroofing of rocks (lines 3, 4, 7, and 21).
6. Transportation of Materials: Erosion involves the movement of soil or rock materials,
either leading to accumulation in new areas or the loss from the original site, as discussed
in lines 13 and 17.
7. Feedback Mechanisms: It can be part of a feedback system where erosion processes
intensify other processes, such as increasing wave attack or causing higher waves, as
mentioned in lines 14 and 22.</p>
        <p>These traits collectively describe erosion as a significant, dynamic process that impacts
landscapes, ecosystems, and human activities, often requiring management or mitigation
to prevent or reduce its detrimental efects.</p>
      </sec>
      <sec id="sec-7-21">
        <title>Example 21: Asking ChatGPT to analyze some concordances lines and then writing a definition based on them</title>
        <p>User
Analyze the following concordance lines of the term ”false seedbed technique” from a semantic
point of view. What semantic traits of the term can you extract from them? Always mention
the lines to justify your answer.
14. The false seedbed technique has not been well documented; however, this practice is
widespread on organic farms (Mohler, 2001) and a reduction in weed density of 63–85%
has been observed in some situations (Gunsolus, 1990; Leblanc and Cloutier, 1996).
15. Reducing weed seed pressure with the false seedbed technique .
16. Despite its limitations, the false seedbed technique has broad applicability.
17. When the false seedbed technique was used (M2), the 5-cm upper soil layer was fragmented
three times before sowing at 5 cm depth with a seedbed cultivator and a power harrow.
18. Our aim was to gain insight into how soil tillage (here, the false seedbed technique ) and
cover crops impact the spontaneous vegetation in fallows preceding banana crops.
19. False seedbed technique aims to reduce weed seed bank by exploiting seed germination
biology.
20. The system utilized for physical weed control was based first on a false seedbed technique
followed by pre-sowing weed removal, performed with a special 2 m wide 6-row
springtine harrow.
21. The false seedbed technique reduces the weed seed bank in the topsoil and, as a result,
significantly reduces competition of annual weeds in the succeeding crop.
22. Moreover, the false seedbed technique may have failed to destroy rhizomes present in the
superficial layers (Lamour and Lotz 2007; Ringselle et al. 2019).
23. The importance of shallow tillage as a weed control method in the false seedbed technique
has been highlighted.
24. False seedbed technique is a method providing weed seed bank depletion.
25. Predicting weed emergence can answer the question of how much time weed control
and crop sowing should be delayed in a specific agricultural area where false seedbed
technique is about to be applied.</p>
        <p>ChatGPT
Analyzing the concordance lines of the term ”false seedbed technique,” several semantic traits
emerge:
1. Alternative Weed Control Method: Lines 1 and 14 indicate that the false seedbed
technique is an alternative to conventional weed control methods, suggesting its use as a
sustainable or less chemically dependent approach.
2. Mechanical Weed Suppression: Lines 5, 7, and 20 describe the technique as primarily
mechanical, involving harrowing or similar tools for seedling destruction, contrasting
with chemical or thermal methods in other techniques.
3. Weed Seed Bank Reduction: Lines 2, 19, 21, and 24 mention the goal of reducing
weed seed pressure or depleting the weed seed bank, indicating the technique’s focus on
preventing weed growth by targeting the seeds before they can establish.
4. Soil Tillage Component: Lines 4, 8, 10, 17, and 23 highlight shallow tillage as a
crucial element of the technique, emphasizing soil manipulation to encourage weed seed
germination before destroying them.
5. Timing and Planning: Lines 9, 10, and 25 suggest that the technique’s success depends
on timing, particularly in relation to weed emergence and tillage operations, to maximize
weed control eficacy.
6. Applicability and Limitations: Lines 12, 14, 16, and 22 discuss the technique’s broad
applicability, especially in organic farming, while also acknowledging its limitations, such
as potential failure to destroy certain weed parts like rhizomes.
7. Integrated Weed Management: Lines 6, 11, and 20 imply that the false seedbed
technique is often part of a broader weed management strategy, integrating mechanical,
cultural, and sometimes chemical methods.</p>
        <p>These traits collectively characterize the false seedbed technique as a proactive, mechanical,
and soil-focused method of weed control, emphasizing the strategic timing and integration
within broader weed management practices.</p>
        <p>User
Write a definition of ”false seedbed technique” ready to be inserted in a terminological resource
based the concordance lines. Include a justification of the definition mentioning the lines on
which you have based your definition.</p>
        <p>ChatGPT
Definition:
The false seedbed technique is a weed management method that involves the shallow tillage
of soil to stimulate the germination of weed seeds, followed by their destruction through
mechanical means such as harrowing or using a spring-tine harrow. This technique aims to
reduce the weed seed bank in the topsoil, thereby decreasing the competition from annual
weeds in subsequent crops.</p>
        <p>Justification:
This definition synthesizes information from several key points in the concordance lines:
• Shallow Tillage and Mechanical Destruction: Lines 4, 5, and 23 emphasize the
importance of shallow tillage as a weed control method within the false seedbed technique
and the use of mechanical tools for seedling destruction.
• Weed Seed Bank Reduction: Lines 2, 19, 21, and 24 specifically mention the objective of
reducing weed seed pressure or depleting the weed seed bank, highlighting the technique’s
preventive approach to weed control.
• Weed Management Strategy: Lines 1, 14, and 16 illustrate the technique’s role as an
alternative weed control method, often utilized in sustainable or organic farming contexts
due to its mechanical nature and minimal soil disturbance.</p>
        <p>This comprehensive view from the concordance lines ensures the definition captures the essence
of the false seedbed technique as a proactive, mechanical weed management practice.</p>
      </sec>
      <sec id="sec-7-22">
        <title>Example 22: Asking ChatGPT to evaluate a definition in search of diferent types of errors</title>
        <p>User
Evaluate this definition in search of the following issues: 1) spelling mistakes, 2) poor stylistic
choices, 3) bias, 4) factual errors. Finally, rewrite the definition taking into account your own
suggestions.
nuclear energy: energy derived from the nuckleus of an atom commonly harnessed
through two methods: fission and fusion. Fission involves the splitting of heavy atomic nuclei
into smaller particles, releasing a significant amount of energy. This method is predominantly
used in nuclear power plants to generate electricity. Fusion, on the other hand, involves the
combination of light atomic nuclei to form a heavier nucleus, releasing energy in the process.
Fusion is the power source of the sun and stars but has not yet been commercially realized on
Earth due to the immense conditions required. Nuclear energy is a safe source of low-carbon
power and ofers a solution to energy demands with reduced greenhouse gas emissions.
ChatGPT
1) Spelling Mistakes:
2) Poor Stylistic Choices:</p>
        <p>• ”nuckleus” should be ”nucleus”.
3) Bias:
4) Factual Errors:
• The definition starts directly with ”energy derived from the nucleus of an atom” which
could be prefaced for clarity.
• The phrase ”which a tremendous amount of energy” is awkward and seems incomplete.
• The statement ”Nuclear energy is a safe source of low-carbon power” presents a positive
view of nuclear energy without acknowledging the associated risks and controversies.
• Fission involves the splitting of *heavy* atomic nuclei, not light atomic nuclei.
• Fusion involves the combination of *light* atomic nuclei, not heavy atomic nuclei, to
form a heavier nucleus.</p>
        <p>Rewritten Definition:
Nuclear energy refers to the power generated from the nucleus of an atom, primarily through
two processes: fission and fusion. Fission is the process of splitting heavy atomic nuclei into
smaller particles, releasing a significant amount of energy. This method is predominantly used
in nuclear power plants to generate electricity. Fusion, conversely, is the merging of light atomic
nuclei to create a heavier nucleus, a process that liberates energy and powers the sun and
other stars. Although fusion promises a substantial energy yield, it has not been commercially
harnessed on Earth due to the extreme conditions required for its initiation. While nuclear
energy contributes to low-carbon power generation and can meet substantial energy demands,
its safety and environmental impact are subjects of ongoing debate and research.</p>
      </sec>
      <sec id="sec-7-23">
        <title>Example 23: Asking ChatGPT to guess the defined term and improve the definition</title>
        <p>ChatGPT
The term defined by this definition is ”carbon sequestration.” It refers to the process, activity, or
mechanism that removes carbon dioxide (CO2) from the atmosphere and stores it in a carbon
sink, which can be either natural, like forests and oceans, or artificial, like carbon capture and
storage technology.</p>
      </sec>
    </sec>
  </body>
  <back>
    <ref-list>
      <ref id="ref1">
        <mixed-citation>
          [1]
          <string-name>
            <given-names>D.</given-names>
            <surname>Gouadec</surname>
          </string-name>
          , Terminologie. Constitution des données,
          <source>AFNOR</source>
          ,
          <year>1990</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref2">
        <mixed-citation>
          [2]
          <string-name>
            <given-names>H.</given-names>
            <surname>Suonuuti</surname>
          </string-name>
          , Guide to terminology, Tekniikan Sanastokeskus ry,
          <year>1997</year>
          . URL: http://www. nordterm.net/wiki/en/index.php/Nordterm_8.
        </mixed-citation>
      </ref>
      <ref id="ref3">
        <mixed-citation>
          [3]
          <string-name>
            <given-names>S.</given-names>
            <surname>Pavel</surname>
          </string-name>
          ,
          <string-name>
            <given-names>D.</given-names>
            <surname>Nolet</surname>
          </string-name>
          , Précis de terminologie, Bureau de la traduction,
          <year>2001</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref4">
        <mixed-citation>
          [4]
          <string-name>
            <given-names>R.</given-names>
            <surname>Dubuc</surname>
          </string-name>
          , Manuel pratique de terminologie,
          <source>Linguatech</source>
          ,
          <year>2002</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref5">
        <mixed-citation>
          [5]
          <string-name>
            <surname>ISO</surname>
          </string-name>
          , ISO
          <volume>704</volume>
          :
          <year>2022</year>
          (
          <article-title>Terminology work -</article-title>
          <source>Principles and methods)</source>
          ,
          <year>2022</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref6">
        <mixed-citation>
          [6]
          <string-name>
            <given-names>R.</given-names>
            <surname>Vézina</surname>
          </string-name>
          ,
          <string-name>
            <given-names>X.</given-names>
            <surname>Darras</surname>
          </string-name>
          ,
          <string-name>
            <given-names>J.</given-names>
            <surname>Bédard</surname>
          </string-name>
          ,
          <string-name>
            <given-names>M.</given-names>
            <surname>Lapointe-Giguère</surname>
          </string-name>
          , La rédaction de définitions terminologiques, Ofice québécois de la langue française,
          <year>2009</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref7">
        <mixed-citation>
          [7]
          <string-name>
            <surname>Termcat</surname>
          </string-name>
          , La definició terminològica,
          <source>Eumo</source>
          ,
          <year>2009</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref8">
        <mixed-citation>
          [8]
          <string-name>
            <given-names>R.</given-names>
            <surname>Temmerman</surname>
          </string-name>
          ,
          <article-title>Towards New Ways of Terminology Description: The sociocognitive approach</article-title>
          , John Benjamins,
          <year>2000</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref9">
        <mixed-citation>
          [9]
          <string-name>
            <given-names>A.</given-names>
            <surname>San Martín</surname>
          </string-name>
          , A Flexible Approach to Terminological Definitions: Representing Thematic Variation,
          <source>International Journal of Lexicography</source>
          <volume>35</volume>
          (
          <year>2022</year>
          )
          <fpage>53</fpage>
          -
          <lpage>74</lpage>
          . doi:
          <volume>10</volume>
          .1093/ijl/ ecab013.
        </mixed-citation>
      </ref>
      <ref id="ref10">
        <mixed-citation>
          [10]
          <string-name>
            <given-names>V.</given-names>
            <surname>Evans</surname>
          </string-name>
          ,
          <string-name>
            <surname>Cognitive Linguistics</surname>
            :
            <given-names>A Complete</given-names>
          </string-name>
          <string-name>
            <surname>Guide</surname>
          </string-name>
          , Edinburgh University Press,
          <year>2019</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref11">
        <mixed-citation>
          [11]
          <string-name>
            <surname>I. Kecskes</surname>
          </string-name>
          , Intercultural Pragmatics, Oxford University Press,
          <year>2014</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref12">
        <mixed-citation>
          [12]
          <string-name>
            <given-names>J.</given-names>
            <surname>Allwood</surname>
          </string-name>
          ,
          <article-title>Meaning potentials and context: Some consequences for the analysis of variation in meaning</article-title>
          , in: H.
          <string-name>
            <surname>Cuyckens</surname>
            ,
            <given-names>R.</given-names>
          </string-name>
          <string-name>
            <surname>Dirven</surname>
          </string-name>
          , J. Taylor (Eds.), Cognitive Approaches to Lexical Semantics, Mouton de Gruyter,
          <year>2003</year>
          , pp.
          <fpage>29</fpage>
          -
          <lpage>66</lpage>
          . doi:
          <volume>10</volume>
          .1515/9783110219074.29.
        </mixed-citation>
      </ref>
      <ref id="ref13">
        <mixed-citation>
          [13]
          <string-name>
            <given-names>V.</given-names>
            <surname>Evans</surname>
          </string-name>
          ,
          <source>How Words Mean: Lexical Concepts</source>
          ,
          <source>Cognitive Models, and Meaning Construction</source>
          , Oxford University Press,
          <year>2009</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref14">
        <mixed-citation>
          [14]
          <string-name>
            <given-names>A.</given-names>
            <surname>San</surname>
          </string-name>
          <string-name>
            <surname>Martín</surname>
          </string-name>
          ,
          <article-title>La representación de la variación contextual mediante definiciones terminológicas flexibles</article-title>
          ,
          <source>Ph.D. thesis</source>
          , University of Granada,
          <year>2016</year>
          . doi: 10481/43423.
        </mixed-citation>
      </ref>
      <ref id="ref15">
        <mixed-citation>
          [15]
          <string-name>
            <given-names>A.</given-names>
            <surname>San</surname>
          </string-name>
          <string-name>
            <surname>Martín</surname>
          </string-name>
          ,
          <article-title>Contextual Constraints in Terminological Definitions, Frontiers in Communication 7 (</article-title>
          <year>2022</year>
          ). doi:
          <volume>10</volume>
          .3389/fcomm.
          <year>2022</year>
          .
          <volume>885283</volume>
          .
        </mixed-citation>
      </ref>
      <ref id="ref16">
        <mixed-citation>
          [16]
          <string-name>
            <given-names>W.</given-names>
            <surname>Croft</surname>
          </string-name>
          ,
          <string-name>
            <given-names>A.</given-names>
            <surname>Cruse</surname>
          </string-name>
          , Cognitive linguistics, Cambridge University Press,
          <year>2004</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref17">
        <mixed-citation>
          [17]
          <string-name>
            <surname>G.-M. de Schryver</surname>
          </string-name>
          ,
          <string-name>
            <surname>Generative</surname>
            <given-names>AI</given-names>
          </string-name>
          and
          <article-title>Lexicography: The Current State of the Art Using ChatGPT</article-title>
          ,
          <source>International Journal of Lexicography</source>
          (
          <year>2023</year>
          ). doi:
          <volume>10</volume>
          .1093/ijl/ecad021.
        </mixed-citation>
      </ref>
      <ref id="ref18">
        <mixed-citation>
          [18]
          <string-name>
            <given-names>A.</given-names>
            <surname>Borji</surname>
          </string-name>
          ,
          <string-name>
            <given-names>Q.</given-names>
            <surname>Ai</surname>
          </string-name>
          ,
          <string-name>
            <surname>A Categorical</surname>
          </string-name>
          <article-title>Archive of ChatGPT Failures, ArXiv preprint (</article-title>
          <year>2023</year>
          ). arXiv:
          <volume>2302</volume>
          .
          <fpage>03494</fpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref19">
        <mixed-citation>
          [19] OpenAI, GPT-4
          <source>Technical Report, Technical Report, OpenAI</source>
          ,
          <year>2023</year>
          . URL: https://arxiv.org/ abs/2303.08774.
        </mixed-citation>
      </ref>
      <ref id="ref20">
        <mixed-citation>
          [20]
          <string-name>
            <given-names>M.</given-names>
            <surname>Jakubíček</surname>
          </string-name>
          ,
          <string-name>
            <given-names>M.</given-names>
            <surname>Rundell</surname>
          </string-name>
          ,
          <article-title>The end of lexicography? Can ChatGPT outperform current tools for post-editing lexicography?</article-title>
          ,
          <source>in: Proceedings of the eLex 2023 conference, Lexical Computing CZ s.r.o.</source>
          ,
          <year>2023</year>
          , pp.
          <fpage>518</fpage>
          -
          <lpage>33</lpage>
          . URL: https://elex.link/elex2023/proceedings-download.
        </mixed-citation>
      </ref>
      <ref id="ref21">
        <mixed-citation>
          [21]
          <string-name>
            <given-names>P.</given-names>
            <surname>Li</surname>
          </string-name>
          ,
          <string-name>
            <given-names>J.</given-names>
            <surname>Yang</surname>
          </string-name>
          ,
          <string-name>
            <given-names>M. A.</given-names>
            <surname>Islam</surname>
          </string-name>
          ,
          <string-name>
            <given-names>S.</given-names>
            <surname>Ren</surname>
          </string-name>
          ,
          <string-name>
            <surname>Making AI</surname>
          </string-name>
          <article-title>Less ”thirsty”: Uncovering and Addressing the Secret Water Footprint of AI Models, ArXiv preprint (</article-title>
          <year>2023</year>
          ). arXiv:
          <volume>2304</volume>
          .
          <fpage>03271</fpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref22">
        <mixed-citation>
          [22]
          <string-name>
            <given-names>N.</given-names>
            <surname>Lucchi</surname>
          </string-name>
          ,
          <source>ChatGPT: A Case Study on Copyright Challenges for Generative Artificial Intelligence Systems</source>
          ,
          <source>European Journal of Risk Regulation</source>
          (
          <year>2023</year>
          )
          <fpage>1</fpage>
          -
          <lpage>23</lpage>
          . doi:
          <volume>10</volume>
          .1017/err.
          <year>2023</year>
          .
          <volume>59</volume>
          .
        </mixed-citation>
      </ref>
      <ref id="ref23">
        <mixed-citation>
          [23]
          <string-name>
            <given-names>M.</given-names>
            <surname>Jakubíček</surname>
          </string-name>
          ,
          <string-name>
            <given-names>M.</given-names>
            <surname>Měchura</surname>
          </string-name>
          ,
          <string-name>
            <given-names>V.</given-names>
            <surname>Kovář</surname>
          </string-name>
          ,
          <string-name>
            <given-names>P.</given-names>
            <surname>Rychlý</surname>
          </string-name>
          ,
          <article-title>Practical Post-Editing Lexicography with Lexonomy and Sketch Engine</article-title>
          ,
          <source>in: Proceedings of the XVIII EURALEX International Congress</source>
          ,
          <year>2018</year>
          , pp.
          <fpage>65</fpage>
          -
          <lpage>67</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref24">
        <mixed-citation>
          [24]
          <string-name>
            <given-names>J.</given-names>
            <surname>Lee</surname>
          </string-name>
          ,
          <string-name>
            <given-names>T.</given-names>
            <surname>Le</surname>
          </string-name>
          ,
          <string-name>
            <given-names>J.</given-names>
            <surname>Chen</surname>
          </string-name>
          ,
          <string-name>
            <given-names>D.</given-names>
            <surname>Lee</surname>
          </string-name>
          , Do Language Models Plagiarize?,
          <source>in: Proceedings of the ACM Web Conference</source>
          <year>2023</year>
          , ACM,
          <year>2023</year>
          , pp.
          <fpage>3637</fpage>
          -
          <lpage>3647</lpage>
          . doi:
          <volume>10</volume>
          .1145/3543507.3583199.
        </mixed-citation>
      </ref>
      <ref id="ref25">
        <mixed-citation>
          [25]
          <string-name>
            <given-names>E.</given-names>
            <surname>Mckean</surname>
          </string-name>
          ,
          <string-name>
            <given-names>W.</given-names>
            <surname>Fitzgerald</surname>
          </string-name>
          ,
          <article-title>The ROI of AI in Lexicography</article-title>
          , in: Asialex 2023 Proceedings,
          <year>2023</year>
          , pp.
          <fpage>18</fpage>
          -
          <lpage>27</lpage>
          . URL: https://www.asialex.org/pdf/Asialex-Proceedings-
          <year>2023</year>
          .pdf.
        </mixed-citation>
      </ref>
      <ref id="ref26">
        <mixed-citation>
          [26]
          <string-name>
            <given-names>R.</given-names>
            <surname>Lew</surname>
          </string-name>
          ,
          <article-title>ChatGPT as a COBUILD lexicographer</article-title>
          ,
          <source>Humanities and Social Sciences Communications</source>
          <volume>10</volume>
          (
          <year>2023</year>
          )
          <article-title>704</article-title>
          . doi:
          <volume>10</volume>
          .1057/s41599-023-02119-6.
        </mixed-citation>
      </ref>
      <ref id="ref27">
        <mixed-citation>
          [27]
          <string-name>
            <given-names>P.</given-names>
            <surname>Faber</surname>
          </string-name>
          ,
          <string-name>
            <given-names>C. I. L.</given-names>
            <surname>Rodríguez</surname>
          </string-name>
          ,
          <string-name>
            <given-names>M.</given-names>
            <surname>Tercedor</surname>
          </string-name>
          , Utilización de técnicas de corpus
          <source>en la representación del conocimiento médico, Terminology</source>
          <volume>7</volume>
          (
          <year>2001</year>
          )
          <fpage>167</fpage>
          -
          <lpage>197</lpage>
          . doi:
          <volume>10</volume>
          .1075/term.7. 2.04fab.
        </mixed-citation>
      </ref>
      <ref id="ref28">
        <mixed-citation>
          [28]
          <string-name>
            <given-names>P.</given-names>
            <surname>Faber</surname>
          </string-name>
          ,
          <string-name>
            <given-names>R. M.</given-names>
            <surname>Usón</surname>
          </string-name>
          ,
          <article-title>Constructing a Lexicon of English Verbs</article-title>
          , Mouton de Gruyter,
          <year>1999</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref29">
        <mixed-citation>
          [29]
          <string-name>
            <given-names>P.</given-names>
            <surname>Faber</surname>
          </string-name>
          ,
          <article-title>Terminographic definition and concept representation</article-title>
          , in: B.
          <string-name>
            <surname>Maia</surname>
            ,
            <given-names>J.</given-names>
          </string-name>
          <string-name>
            <surname>Haller</surname>
          </string-name>
          , M. Ulyrich (Eds.),
          <source>Training the Language Services Provider for the New Millennium, Universidade do Porto</source>
          ,
          <year>2002</year>
          , pp.
          <fpage>343</fpage>
          -
          <lpage>354</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref30">
        <mixed-citation>
          [30]
          <string-name>
            <given-names>A.</given-names>
            <surname>Kilgarrif</surname>
          </string-name>
          ,
          <string-name>
            <given-names>P.</given-names>
            <surname>Rychly</surname>
          </string-name>
          ,
          <string-name>
            <given-names>P.</given-names>
            <surname>Smrz</surname>
          </string-name>
          ,
          <string-name>
            <given-names>D.</given-names>
            <surname>Tugwell</surname>
          </string-name>
          , The Sketch Engine, in: G. Williams,
          <string-name>
            <surname>S.</surname>
          </string-name>
          Vessier (Eds.),
          <source>Proceedings of the Eleventh EURALEX International Congress, EURALEX</source>
          ,
          <year>2004</year>
          , pp.
          <fpage>105</fpage>
          -
          <lpage>116</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref31">
        <mixed-citation>
          [31]
          <string-name>
            <given-names>P.</given-names>
            <surname>León-Araúz</surname>
          </string-name>
          ,
          <string-name>
            <surname>A</surname>
          </string-name>
          . San Martín,
          <article-title>The EcoLexicon Semantic Sketch Grammar: from Knowledge Patterns to Word Sketches</article-title>
          , in: I.
          <string-name>
            <surname>Kerneman</surname>
          </string-name>
          , S. Krek (Eds.),
          <source>Proceedings of the LREC 2018 Workshop “</source>
          Globalex 2018 - Lexicography &amp; WordNets”, Globalex,
          <year>2018</year>
          , pp.
          <fpage>94</fpage>
          -
          <lpage>99</lpage>
          . URL: http://lrec-conf.org/workshops/lrec2018/W33/pdf/15_W33.pdf.
        </mixed-citation>
      </ref>
      <ref id="ref32">
        <mixed-citation>
          [32]
          <string-name>
            <given-names>A.</given-names>
            <surname>San Martín</surname>
          </string-name>
          , C. Trekker,
          <string-name>
            <given-names>P.</given-names>
            <surname>León-Araúz</surname>
          </string-name>
          ,
          <article-title>Repérage automatisé de l'hyponymie dans des corpus spécialisés en français à l'aide de Sketch Engine</article-title>
          ,
          <source>Terminology</source>
          <volume>28</volume>
          (
          <year>2022</year>
          )
          <fpage>264</fpage>
          -
          <lpage>298</lpage>
          . doi:
          <volume>10</volume>
          .1075/term.20044.san.
        </mixed-citation>
      </ref>
      <ref id="ref33">
        <mixed-citation>
          [33]
          <string-name>
            <given-names>M.</given-names>
            <surname>Rundell</surname>
          </string-name>
          ,
          <article-title>Automating the creation of dictionaries: are we nearly there?</article-title>
          ,
          <source>in: Asialex 2023 Proceedings</source>
          ,
          <year>2023</year>
          , pp.
          <fpage>9</fpage>
          -
          <lpage>17</lpage>
          . URL: https://www.asialex.org/pdf/Asialex-Proceedings-
          <year>2023</year>
          . pdf.
        </mixed-citation>
      </ref>
      <ref id="ref34">
        <mixed-citation>
          1.
          <article-title>Radiative forcing occurs at the top of the troposphere, specifically at the tropopause</article-title>
          .
          <source>(EIA</source>
          , IPCC) Accurate
        </mixed-citation>
      </ref>
      <ref id="ref35">
        <mixed-citation>
          2.
          <article-title>It involves a change in average net radiation due to variations in incoming solar or outgoing infrared radiation</article-title>
          .
          <source>(EIA</source>
          , IPCC) Accurate
        </mixed-citation>
      </ref>
      <ref id="ref36">
        <mixed-citation>
          3.
          <article-title>Positive radiative forcing tends to warm the Earth's surface, while negative radiative forcing tends to cool it. (EIA, CARA</article-title>
          , OSU) Accurate
        </mixed-citation>
      </ref>
      <ref id="ref37">
        <mixed-citation>
          4.
          <article-title>Greenhouse gases trap infrared energy radiated from the Earth's surface, leading to positive radiative forcing. (EIA, UNFCCC</article-title>
          , EO-NASA) Accurate
        </mixed-citation>
      </ref>
      <ref id="ref38">
        <mixed-citation>
          5.
          <article-title>Radiative forcing afects the balance between incoming solar radiation and outgoing infrared radiation. (UNFCCC, CARA, EO-NASA, IPCC</article-title>
          , NAL) Accurate
        </mixed-citation>
      </ref>
      <ref id="ref39">
        <mixed-citation>
          6.
          <article-title>Without radiative forcing, solar radiation coming to Earth would be approximately equal to the infrared radiation emitted from Earth. (UNFCCC, CARA, EO-NASA, NAL</article-title>
          ) Accurate
        </mixed-citation>
      </ref>
      <ref id="ref40">
        <mixed-citation>
          7.
          <article-title>Causes of radiative forcing include internal changes and external factors like solar output or carbon dioxide concentrations</article-title>
          .
          <source>(CARA) Accurate</source>
        </mixed-citation>
      </ref>
      <ref id="ref41">
        <mixed-citation>
          8.
          <article-title>Greenhouse gases increase the fraction of trapped infrared radiation, reradiating it back toward the Earth's surface, creating a warming efect. (UNFCCC, EONASA</article-title>
          , NAL) Accurate
        </mixed-citation>
      </ref>
      <ref id="ref42">
        <mixed-citation>
          9.
          <article-title>It is measured in watts per square meter</article-title>
          (W/m2̂).
          <source>(IPCC</source>
          ,
          <string-name>
            <surname>Climate</surname>
            <given-names>CoLab</given-names>
          </string-name>
          , GCMDNASA, CCCR, OSU) Accurate
        </mixed-citation>
      </ref>
      <ref id="ref43">
        <mixed-citation>
          10.
          <article-title>Radiative forcing reflects changes in the energy balance of the earth-atmosphere system due to various factors such as greenhouse gases and solar radiation. (Climate CoLab</article-title>
          , GCMD-NASA) Accurate
        </mixed-citation>
      </ref>
      <ref id="ref44">
        <mixed-citation>
          11.
          <article-title>The climate system inherently tries to balance incoming and outgoing radiation. (Climate CoLab) Accurate</article-title>
        </mixed-citation>
      </ref>
    </ref-list>
  </back>
</article>