=Paper=
{{Paper
|id=Vol-3834/paper60
|storemode=property
|title=SCIENCE IS EXPLORATION: Computational Frontiers for Conceptual Metaphor Theory
|pdfUrl=https://ceur-ws.org/Vol-3834/paper60.pdf
|volume=Vol-3834
|authors=Rebecca M. M. Hicke,Ross Deans Kristensen-McLachlan
|dblpUrl=https://dblp.org/rec/conf/chr/HickeK24
}}
==SCIENCE IS EXPLORATION: Computational Frontiers for Conceptual Metaphor Theory==
https://ceur-ws.org/Vol-3834/paper60.pdf
SCIENCE IS EXPLORATION: Computational
Frontiers for Conceptual Metaphor Theory
Rebecca M. M. Hicke1,2,∗ , Ross Deans Kristensen-McLachlan2,3
1
Department of Computer Science, Cornell University, USA
2
Center for Humanities Computing, Aarhus University, Denmark
3
Department of Linguistics, Cognitive Science, and Semiotics, Aarhus University, Denmark
Abstract
Metaphors are everywhere. They appear extensively across all domains of natural language, from the
most sophisticated poetry to seemingly dry academic prose. A significant body of research in the cogni-
tive science of language argues for the existence of conceptual metaphors, the systematic structuring of
one domain of experience in the language of another. Conceptual metaphors are not simply rhetorical
flourishes but are crucial evidence of the role of analogical reasoning in human cognition. In this paper,
we ask whether Large Language Models (LLMs) can accurately identify and explain the presence of such
conceptual metaphors in natural language data. Using a novel prompting technique based on metaphor
annotation guidelines, we demonstrate that LLMs are a promising tool for large-scale computational
research on conceptual metaphors. Further, we show that LLMs are able to apply procedural guidelines
designed for human annotators, displaying a surprising depth of linguistic knowledge.
Keywords
conceptual metaphor theory, large language models, pragglejaz
”Metaphors are much more tenacious than facts.”
- Paul de Man [19]
1. Introduction
Metaphor is commonly understood as the description of one concept in the vocabulary of an-
other, typically for poetic, rhetorical, or otherwise literary effect. However, some metaphors
are so systematic and conventionalized that we barely recognize them as such. Consider, for
example, the commonplace metaphor life is a journey. While this is certainly a metaphor in
its own right, it can also be instantiated in ways which do not directly refer to either life or
journeys: ”We have come a long way”; ”I’m going through a rough patch”; ”She’s at a crossroads.”
A metaphor like life is a journey is more than simple figurative language. Instead, it allows
us to make sense of the process of a normal human life through the rich language of journeys
and voyages. These kinds of systematic mappings are known as conceptual metaphors [16].
CHR 2024: Computational Humanities Research Conference, December 4–6, 2024, Aarhus, Denmark
∗
Corresponding author.
£ rmh327@cornell.edu (R. M. M. Hicke); rdkm@cc.au.dk (R. D. Kristensen-McLachlan)
ç https://rmatouschekh.github.io (R. M. M. Hicke)
ȉ 0009-0006-2074-8376 (R. M. M. Hicke); 0000-0001-8714-1911 (R. D. Kristensen-McLachlan)
© 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
1105
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
Conceptual metaphors are so deeply ingrained in our everyday language that the most fre-
quent usage of a word or phrase may be metaphorical. This can make it difÏcult to recognize
the presence of metaphorical language, even for linguists and other specialists [14]. Moreover,
processing and understanding conceptual metaphors necessarily involves encyclopedic general
knowledge of the world, as well as shared cultural norms and stereotypes between speaker/-
hearer or writer/reader. Identifying conceptual metaphors has therefore consistently proven to
be a challenge for both natural language processing (NLP) and artificial intelligence (AI). Given
that contemporary transformer-based models ‘learn’ by consuming huge amounts of natural
language and build knowledge by looking at how words are commonly used, it seems that they
should similarly struggle to recognize these metaphors.
Nevertheless, being able to consistently identify the presence of conceptual metaphors in dis-
course would be greatly beneficial for fields such as computational literary studies and cultural
analytics more broadly. In this paper, we therefore set out to test empirically whether LLMs
can be used productively in the context of conceptual metaphor theory (CMT). If the answer
is no, it may be evidence that developing computational approaches to conceptual metaphor
remains an intractable problem. If, on the other hand, LLMs can consistently identify the pres-
ence of conceptual metaphors, the question then arises of exactly how and where LLMs encode
metaphoricity.
2. Related Work
Cognitive Linguistic Foundations
The study of CMT was inaugurated in the 1980s in Lakoff and Johnson [16]. Central to the ar-
gument in [16] is the fundamental role that metaphor plays in structuring everyday discourse.
Rather than simply being a literary or rhetorical flourish, Lakoff and Johnson argue that concep-
tual metaphors structure our common understanding of entities, processes, or entire semantic
domains. CMT played a crucial role in the development of non-generativist cognitive linguis-
tics [7, 17, 18, 27, 28]. An important development came with the introduction of conceptual
blending (otherwise known as conceptual integration), in which the kinds of analogical reason-
ing underlying conceptual metaphors are argued to be more fundamental to human language
and cognition [11, 12, 8]. While there are dissenting voices [13], recent decades have seen a
growing body of empirical research which claims to provide experimental evidence in favour
of CMT [4, 6, 29].
Metaphor and Artificial Intelligence
The extensive role and function of metaphor in natural language has not escaped researchers
in NLP. Indeed, a great deal of early work in AI concerned exactly how a computer could make
sense of metaphorical language [5, 23, 24] (see [2] for a comprehensive overview). Much of this
work relied on so-called symbolic AI, but struggled because GOFAI (good, old-fashioned AI)
could not provide the analogous reasoning skills and extensive general knowledge required
to process metaphorical language. In the context of conceptual blending specifically, there
have been tentative attempts to model the underlying processes involved computationally [10].
1106
1. Read the entire text-discourse to establish a general
understanding of the meaning.
2. Determine the lexical units in the text-discourse.
3.
a. For each lexical unit in the text, establish its
meaning in context, that is, how it applies to an
entity, relation, or attribute in the situation evoked
by the text (contextual meaning). Take into account
what comes before and after the lexical unit.
b. For each lexical unit, determine if it has a more
basic contemporary meaning in other contexts than the
one in the given context. For our purposes, basic
meanings tend to be
i. More concrete; what they evoke is easier to
imagine, see, hear, feel, smell, and taste.
ii. Related to bodily action.
iii. More precise (as opposed to vague).
iv. Historically older.
Basic meanings are not necessarily the most frequent
meanings of the lexical unit.
c. If the lexical unit has a more basic current-
contemporary meaning in other contexts than the given
context, decide whether the contextual meaning
contrasts with the basic meaning but can be understood
in comparison with it.
4. If yes, mark the lexical unit as metaphorical.
Figure 1: The metaphor identification procedure (MIP) introduced by the Pragglejaz Group in [14].
However, the encyclopedic world knowledge required in CMT is challenging to encode since
the domain is essentially unbounded. Moreover, a core issue at the heart of computational
metaphor analysis has historically been the lack of a shared definition of what is being studied,
along with the scarcity of robust data and evaluation strategies [26].
Recent Developments
However, recent years have again seen a growing interest in quantitative and computational
approaches to metaphor, primarily in more applied fields such as corpus linguistics and stylis-
tics [9, 15, 25]. With the advent of neural language models, research in NLP has increasingly
attempted to discover if (and how) word embeddings capture aspects of metaphoricity [20, 30,
21, 22]. More recently, work has been done on explicitly probing pre-trained language mod-
els to inspect how metaphorical knowledge is encoded [1] and creating shared task datasets
[30]. In the context of LLMs specifically, GPT-3 has been shown to perform reasonably well
at predicting source domains for English metaphors, although it performs significantly worse
for Spanish [31]. These are promising developments and point to the increasing awareness of
the practical and theoretical challenges underlying the nature of figurative language in NLP.
To date, though, we are unaware of existing research which seeks to operationalize existing
metaphor identification procedures for LLMs.
1107
For each word in the following text, determine if it has a
*more* basic contemporary meaning in other contexts than the
one in the given context. For our purposes, basic meanings
tend to be:
- more concrete; what they evoke is easier to imagine, see,
here, feel, smell, and taste
- related to bodily action
- more precise (as opposed to vague)
- historically older
Basic meanings are not necessarily the most frequent meanings
of the word. Remember to only report YES for words whose use
in this context is not their most basic meaning.
Examples:
“I demolished his argument.”
- I: NO (pronoun, basic meaning is the same)
- demolished: YES (verb, more basic meaning refers to
destroying a building)
- his: NO (possessive pronoun, basic meaning is the same)
- argument: NO (noun, basic meaning is the same)
“They write about their family.”
- They: NO (pronoun, basic meaning is the same)
- write: NO (verb, basic meaning is the same)
- about: NO (preposition, basic meaning is the same)
- their: NO (possessive pronoun, basic meaning is the same)
- family: NO (noun, basic meaning is the same)
Figure 2: The modified prompt used in all experiments. All significant (non-formatting) changes from
Step 3b of the MIP procedure are highlighted in red. The sentence to be analyzed is appended in
quotations to the bottom of the prompt when querying models.
3. Methods
We are interested in studying whether LLMs are capable of leveraging the contextual and cul-
tural knowledge necessary to perform difÏcult metaphor identification tasks, such as recogniz-
ing conceptual metaphors. To do this, we operationalize the Metaphor Identification Procedure
(MIP), a set of annotation guidelines introduced by the Pragglejaz Group in 2007 [14]. The steps
of MIP are detailed in Figure 1. MIP and its variants are among the most commonly used pro-
cedures for metaphor annotation in work in corpus linguistics and stylistics; we therefore take
it as a reasonable proxy for the metaphor identification process in humans.
Examining MIP, the first few steps have clear parallels in transformer-based LLMs. Step 1
is performed by the attention mechanism. Creating the representation of each word, in tradi-
tional attention mechanisms, requires ‘looking’ at every other word in the text, thus allowing
the model to “establish a general understanding of [its] meaning.” Step 2, determining the
lexical units in a text, is directly paralleled by tokenization. Finally, Step 3a is achieved via
contextual embeddings, whose explicit purpose is to represent the specific meaning of a word
in context. However, Step 3b of MIP has no clear counterpart in transformer-based models.
This is because the ‘basic meaning’ of a lexical unit, as defined by MIP, is not obviously stored
or represented within these models.
We want to know if state-of-the-art LLMs are capable of replicating Step 3b of the MIP
1108
procedure despite these barriers, so we transform it into a prompt for instruction-tuned models
(Figure 2). In creating the prompt, we leave the text of Step 3b almost entirely intact. “Lexical
units” are replaced by “words” in order to encourage consistent outputs. Emphasis is added in
several locations to only annotating words which have a more basic meaning than their usage
in the input text. Finally, two examples are appended to the prompt. These examples are meant
to facilitate in-context learning and provide a structure for the models to mimic in their outputs.
They instruct the model to provide a parenthetical expression next to each word consisting of
the word’s part of speech and its more basic meaning, if one exists. These expressions were
introduced by the gpt-4o model itself during initial exploration and were then included in the
prompt because they force the model to provide explanations for its annotations. This allows
for further evaluation of the outputs and ultimately means they play a similar role to Chain of
Thought prompts, which have been shown to improve model performance across a variety of
tasks [32].
We use this prompt to query three models from OpenAI’s GPT family, chosen because of
their size, popularity, and high performance on a wide range of tasks. Specifically, we look at
gpt-3.5-turbo,1 gpt-4-turbo,2 and gpt-4o.3 We access each model using OpenAI’s Chat
Completion API. A system prompt (“You are a helpful assistant. You have extensive linguistic
knowledge.” ) is provided each time the model is queried. Then, the main prompt (Figure 2) is
passed as a user message, with the current text of interest appended in quotations at the end. All
parameters except nuclear sampling (top_p) are left at their default values during prompting
for all models. We set top_p to 0.1, which means that the models only select from the tokens
which make up the top 10% of the probability mass when choosing each next token. We adjust
the top_p value because we are not prompting for diverse or creative text, but instead for
reliable and accurate outputs, which we hypothesize will be more likely produced from high-
probability tokens. The cost of all experiments using the OpenAI API was ∼$100.
4. Data
We evaluate each model on two datasets. First, we wish to study the models’ basic ability
to use MIP to correctly identify metaphorical words. To this end, we use the Trope Finder
(TroFi) dataset [3], which consists of sentences drawn from Wall Street Journal issues published
between 1987 and 1989. Each sentence contains one of a list of fifty words whose usage is
annotated as literal or non-literal. We only consider sentences from the dataset that have been
labeled by humans and take their annotations as ground truth, excluding all sentences labeled
only by the TroFi clustering system. This leaves us with 3,736 sentences. Each model is then
prompted as described above for each sentence and evaluated only on the words annotated in
TroFi. A response is marked as correct if the model says “YES” there is a more basic meaning
for the word of interest and the label from TroFi is “literal” and vice versa. Occasionally, the
models provide labels for a phrase in the input sentence instead of individual words; the label
provided for the entire phrase is then applied to the word of interest if necessary. If a model
1
gpt-3.5-turbo-0125
2
gpt-4-turbo-2024-04-09
3
gpt-4o-2024-05-13
1109
Table 1
The precision and recall values for the identification of literal and metaphorical word usage in the TroFi
dataset by all models.
Metaphorical Literal
Precision Recall Precision Recall
3.5-turbo 58.30 97.90 66.42 5.59
4-turbo 74.80 86.90 77.41 60.53
4o 73.40 93.66 83.69 54.24
does not provide a label for a word, we act as if the incorrect label was given.
However, the TroFi dataset does not permit for a very fine-grained analysis of the models’
capabilities. In order to facilitate such an analysis, we create a new dataset consisting of exam-
ple sentences pulled from Lakoff and Johnson’s original work on CMT [16]. We collect all full
sentence examples from the text that we believe can be understood as metaphorical without
context, leaving us with 544 sentences, which we refer to as the MWLB dataset. The dataset is
made publicly available, along with all model outputs and human annotations.4
Each model is again prompted to provide word-level annotations for every sentence. Then,
we perform detailed qualitative analysis on each model’s output for a subset of 100 sentences.
Specifically, we evaluate each output on five binary categories:
• L&J Metaphor(s) – Identified: Whether the model has correctly identified all
metaphors highlighted by Lakoff and Johnson. For metaphorical phrases, the model’s
response is considered correct if it has identified at at least one key word as having a
more basic meaning.
• L&J Metaphor(s) – Correct Basic Meanings: Whether the model has provided a cor-
rect, or plausible, more basic meaning for all metaphorical words identified in the cate-
gory above. A label of 1 is applied only if all metaphors from Lakoff and Johnson have
been correctly identified.
• Additional Annotations: Whether the model has labeled any words not highlighted
by Lakoff and Johnson as having a more basic meaning.
• Additional Annotations – Metaphorical: Whether all the additional words labeled
by the model are plausibly metaphorical or may have a more basic meaning.
• Additional Annotations – Correct Basic Meanings: Whether the model has pro-
vided a correct, or plausible, basic meaning for all additional labeled words. It is pos-
sible for models to give the correct basic meaning for a word even if it is not plausibly
metaphorical.
All annotations are performed by one of the authors, using MIP as a guideline. Confusing
or difÏcult annotations are then discussed between both authors till a consensus is reached.
These annotations are necessarily subjective, and we do not claim that they represent a ground
truth. Instead, since these evaluations are informed by the same procedure used to create the
4
https://github.com/rmatouschekh/science-is-exploration
1110
TroFi Dataset Model Confusion Matrices
Literal
3.5-turbo 4-turbo 4o
Literal
Literal
1502 89 628 963 728 863
3000
2000
Metaphorical
Metaphorical
Metaphorical
1000
2100 45 1864 281 2009 136
0
Metaphorical Literal Metaphorical Literal Metaphorical Literal
Figure 3: Confusion matrices for results on the TroFi dataset for each model. The rows represent the
true values for each sample, and the columns represent the model labels.
outputs, we claim that they allow us to study the plausibility of the models’ performance, if
not strictly their accuracy. In future work, further evaluation of model outputs with multiple
trained annotators may allow for a more concrete analysis of model performance.
5. Results and Discussion
On the TroFi dataset, all models demonstrate an ability to distinguish between literal and non-
literal word usage with generally better than chance performance (Table 1). Thus, they seem to
be able to apply the MIP procedure for identifying metaphors and determine computationally
whether a word has a more basic meaning. This is surprising since, as discussed above, basic
word meanings are not necessarily the most frequent. Nothing about the language modeling
approach to pre-training models guarantees that a model would be able to differentiate between
the most frequent and most basic meaning of a word.
We also find, however, that the models are prone to over-labeling words as metaphorical
(Figure 3). Whereas all three models achieve high recall for metaphor identification, they fre-
quently label words used literally as having a more basic meaning, leading to low recall for
the ‘literal’ category and low precision for the ‘metaphorical’ category. 3.5-turbo particu-
larly struggles with accurately identifying literal word usage. It is worth noting that, although
there are clear cases where the models have failed to properly apply MIP, there may also be a
disconnect between annotations in the TroFi corpus and what is considered metaphorical by
the MIP procedure, as MIP was not used for annotation in TroFi.
For the annotated subset of the MWLB dataset, all the models found all the Lakoff and
Johnson metaphors in over 60% of sentences and further provided a correct basic meaning
for the metaphorical words in over 50% of sentences (Figure 4). 4o performed the best on this
task, achieving a remarkable performance given the subtlety and complexity of many of the
metaphors and the cultural knowledge required to identify both their metaphoricity and their
basic meanings. 4o’s high performance may be due in part to the fact that it was most likely to
label words as metaphorical overall, having annotated additional words in the largest percent-
1111
Correctly Identified Metaphors from Lakoff and Johnson
4o 65% 74%
Model
3.5-turbo 52% 63%
Metaphors
4-turbo 57% 62% Metaphors &
Basic Meanings
0 20 40 60 80 100
% Correctly Labeled
Figure 4: The percentage of samples from Lakoff and Johnson where all metaphors and basic meanings
are correctly identified.
age of sentences (Figure 5). In only 48% of sentences did these additional annotations identify
only plausibly metaphorical words. Nonetheless, the inclusion of correct basic meanings for
such a large proportion of the Lakoff and Johnson metaphors suggests that MIP was overall
being correctly applied and true ‘knowledge’ was being demonstrated, not just random chance.
3.5-turbo struggled the most at replicating the desired output structure, occasionally ex-
cluding words or truncating sentences, inconsistently formatting lists, and sometimes drop-
ping the parenthetical explanations. This affected its accuracy, as the model sometimes failed
to provide annotations for words central to the metaphors of interest. Additionally, 4-turbo
was least likely to annotate a word as having a more basic meaning (Figure 5). This meant
it identified fewer Lakoff and Johnson metaphors, but a greater proportion of the additional
metaphors it annotated were plausible. The basic meanings provided by 4-turbo were usually
of high quality, and tended to be more accurate and specific than those produced by 3.5-turbo.
Despite the benefits, however, this model’s ‘caution’ led it to perform worse than 4o overall.
Each model struggled to correctly identify when smaller function words were used metaphor-
ically, particularly prepositions like ‘in,’ . This made the models worse at identifying so-called
container metaphors, such as life is a container and activities are containers. For exam-
ple, in the sentence “That’s in the center of my field of vision.”, labeled by Lakoff and Johnson
as visual fields are containers, the word ‘in’ was overlooked as metaphorical by all three
models.
The models were also challenged by metaphors in which an entity is being treated as a differ-
ent kind of object, like instances of personification, place for institution, or producer for
product metaphors. However, the GPT-4 models had a much greater ability to detect and pro-
vide accurate basic meanings for these. For example, in “Let’s not let Thailand become another
Vietnam.”, both 4-turbo and 4o correctly identified ‘Vietnam’ as a metaphor. 4o explained that
the more basic meaning of the word “refers to the country, whereas in this context it refers to a
situation similar to the Vietnam War” and 4-turbo provides a similar response. Likewise, in“I
hate to read Heidegger.”, 4o recognized that ‘Heidegger’ is being used metaphorically and stated
that the “more basic meaning refers to the person Martin Heidegger, a German philosopher,
rather than his works.” Identifying and explaining both of these metaphors requires a nuanced
understanding of both the semantics of the sentence and the cultural context surrounding them.
1112
% of Samples with Additional Metaphors Labeled
85%
4o 54%
48%
Model
89%
3.5-turbo 45%
40% Legend
% Correct Basic Meanings
91%
4-turbo 34% % Plausibly Metaphorical
65%
0 20 40 60 80 100
% Samples
Figure 5: The percentage of samples from Lakoff and Johnson in which additional metaphors have
been labeled. Smaller bars are included which represent the percentage of additional words for which
the correct basic meaning is provided (upper bar) and which are plausible metaphors (lower bar).
The models cannot always perform this analysis (all three miss that ‘the Alamo’ is metaphorical
in “Remember the Alamo!”), but it is impressive that they are ever able to do so.
In addition, forcing the models to annotate word-by-word makes it challenging for them
to identify metaphors comprising multi-word units. For example, it is difÏcult to determine
which words should be marked as having more basic meanings in the sentence “Get the most
out of life.” Analyzing texts word-by-word can also make identifying and providing the basic
meanings of metaphorical compound words more complex. For example, both ‘underage’ and
‘brainchild’ cause problems for 3.5-turbo; it does not recognize that the under of underage
is metaphorical and it says that the more basic meaning of brainchild is “a child conceived
in the mind.” In contrast, 4-turbo says that the “more basic meanings of ‘brain’ and ‘child’
are more concrete and related to physical objects or beings” and notes that the “more basic
meaning of ‘under’ is physically beneath something” for underage. 4o recognizes both words
as metaphorical, but provides worse basic meanings.
While the models are clearly fallible, they nevertheless demonstrate an impressive ability to
synthesize semantic, syntactic, and cultural information. They are frequently able to recognize
when a word is being used metaphorically and often provide a correct basic meaning for the
word. This ability sometimes holds even for nuanced and complex metaphors.
6. Conclusion
We find that large, generative LMs are capable of applying the classic metaphor annotation
procedure, MIP. In doing so, they demonstrate an ability to discern the “basic meaning” of
words and thus a depth of linguistic understanding that is not obviously gleaned from language
modeling pre-training. Notably, they also demonstrate that LLMs are able to execute linguis-
tic procedures designed for human annotators. This capacity means that generative LLMs
are a promising tool for large-scale computational research on conceptual metaphors, which
has previously been largely infeasible. In addition, it suggests that further research on how
1113
metaphoricity is learned by models may provide insight into their ability to acquire complex
linguistic knowledge that often relies on people’s embodied experiences. These findings also
suggest several avenues for future research, including studies into where information about
words’ basic meanings or metaphoricity is stored by models, an exploration of open source
models ability to annotate for conceptual metaphor, and research on operationalizing other
linguistic procedures for execution by LLMs.
Acknowledgments
We would like to thank Lavinia Cerioni for her help in brainstorming this project and our
colleagues at the Center for Humanities Computing for their support and advice. Additionally,
part of the computation done for this project was performed on the UCloud interactive HPC
system, which is managed by the eScience Center at the University of Southern Denmark.
References
[1] E. Aghazadeh, M. Fayyaz, and Y. Yaghoobzadeh. “Metaphors in Pre-Trained Language
Models: Probing and Generalization Across Datasets and Languages”. In: Proceedings of
the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long
Papers). Dublin, Ireland: Association for Computational Linguistics, 2022, pp. 2037–2050.
doi: 10.18653/v1/2022.acl-long.144.
[2] J. A. Barnden. “Metaphor and Artificial Intelligence: Why They Matter to Each Other”. In:
The Cambridge Handbook of Metaphor and Thought. Cambridge University Press, 2008,
pp. 311–338. doi: 10.1017/cbo9780511816802.020.
[3] J. Birke and A. Sarkar. “A Clustering Approach for Nearly Unsupervised Recognition
of Nonliteral Language”. In: 11th Conference of the European Chapter of the Association
for Computational Linguistics. Trento, Italy: Association for Computational Linguistics,
2006, pp. 329–336.
[4] L. Boroditsky. “Metaphoric Structuring: Understanding Time Through Spatial
Metaphors”. In: Cognition 75.1 (2000), pp. 1–28. doi: 10.1016/s0010-0277(99)00073-6.
[5] J. G. Carbonell. “Metaphor: An Inescapable Phenomenon in Natural Language Com-
prehension”. In: Strategies for Natural Language Processing. Pittsburgh, PA: Psychology
Press, 1982, pp. 415–434. doi: 10.4324/9781315802671.
[6] D. Casasanto and L. Boroditsky. “Time in the Mind: Using Space to Think About Time”.
In: Cognition 106.2 (2008), pp. 579–593. doi: 10.1016/j.cognition.2007.03.004.
[7] W. Croft and D. A. Cruse. Cognitive Linguistics. Cambridge, UK: Cambridge University
Press, 2004. doi: 10.1017/cbo9780511803864.
[8] B. Dancygier and E. Sweetser. Mental Spaces in Grammar: Conditional Constructions.
Cambridge Studies in Linguistics. Cambridge, UK: Cambridge University Press, 2005. doi:
10.1017/cbo9780511486760.
1114
[9] A. Deignan and E. Semino. “Corpus Techniques for Metaphor Analysis”. In: Metaphor
Analysis. London, UK, 2010, pp. 161–179.
[10] M. Eppe, E. Maclean, R. Confalonieri, O. Kutz, M. Schorlemmer, E. Plaza, and K.-U. Kühn-
berger. “A Computational Framework for Conceptual Blending”. In: Artificial Intelligence
256 (2018), pp. 105–129. doi: 10.1016/j.artint.2017.11.005.
[11] G. Fauconnier. Mental Spaces: Aspects of Meaning Construction in Natural Language. Cam-
bridge, UK: Cambridge University Press, 1994. doi: 10.1017/cbo9780511624582.
[12] G. Fauconnier and M. Turner. The Way We Think: Conceptual Blending and the Mind’s
Hidden Complexities. New York City, NY: Basic Books, 2002.
[13] R. W. Gibbs. “Why Do Some People Dislike Conceptual Metaphor Theory?” In: Cognitive
Semiotics 5.1 (2009), pp. 14–36. doi: 10.1515/cogsem.2013.5.12.14.
[14] P. Group. “MIP: A Method for Identifying Metaphorically Used Words in Discourse”. In:
Metaphor and Symbol 22.1 (2007), pp. 1–39. doi: 10.1207/s15327868ms2201\_1.
[15] T. Krennmayr. “What Corpus Linguistics Can Tell Us About Metaphor Use In Newspaper
Texts”. In: Journalism Studies 16.4 (2015), pp. 530–546. doi: 10.1080/1461670x.2014.9371
55.
[16] G. Lakoff and M. Johnson. Metaphors We Live By. Chicago: The University of Chicago
Press, 1980. doi: 10.7208/chicago/9780226470993.001.0001.
[17] R. W. Langacker. Foundations of Cognitive Grammar: Volume I: Theoretical Prerequisites.
Stanford, CA: Stanford University Press, 1987.
[18] R. W. Langacker. Foundations of Cognitive Grammar: Volume II: Descriptive Application.
Stanford, CA: Stanford University Press, 1991. doi: 10.1515/9780804764469.
[19] P. de Man. “Semiology and Rhetoric”. In: Diacritics 3.3 (1973), pp. 27–33. doi: 10.2307/46
4524.
[20] R. Mao, C. Lin, and F. Guerin. “Word Embedding and WordNet Based Metaphor Identifi-
cation and Interpretation”. In: Proceedings of the 56th Annual Meeting of the Association
for Computational Linguistics (Volume 1: Long Papers). Melbourne, Australia: Association
for Computational Linguistics, 2018, pp. 1222–1231. doi: 10.18653/v1/p18-1113.
[21] P. V. Panicheva, I. D. Mamaev, and T. A. Litvinova. “Towards Automatic Conceptual
Metaphor Detection for Psychological Tasks”. In: Information Processing & Management
60.2 (2023), pp. 1–17. doi: 10.1016/j.ipm.2022.103191.
[22] M. Ptiček and J. Dobša. “Methods of Annotating and Identifying Metaphors in the Field
of Natural Language Processing”. In: Future Internet 15.6 (2023), p. 201. doi: 10.3390/fi1
5060201.
[23] S. W. Russell. “Computer Understanding of Metaphorically Used Verbs”. In: American
Journal of Computational Linguistics 3.3 (1976), pp. 1–74.
[24] S. W. Russell. “Information and Experience in Metaphor: A Perspective From Computer
Analysis”. In: Metaphor and Symbolic Activity 1.4 (1986), pp. 227–270. doi: 10.1207/s153
27868ms0104\_1.
1115
[25] E. Semino, Z. Demjén, A. Hardie, S. Payne, and P. Rayson. Metaphor, Cancer and the End
of Life: A Corpus-Based Study. Vol. 19. Routledge Advances in Corpus Linguistics. Oxford,
UK: Routledge, 2018. doi: 10.4324/9781315629834.
[26] E. Shutova. “Design and Evaluation of Metaphor Processing Systems”. In: Computational
Linguistics 41.4 (2015), pp. 579–623. doi: 10.1162/coli\_a\_00233.
[27] L. Talmy. Toward a Cognitive Semantics, Volume 1: Concept Structuring Systems. Cam-
bridge, MA: The MIT Press, 2000. doi: 10.7551/mitpress/6847.001.0001.
[28] L. Talmy. Toward a Cognitive Semantics, Volume 2: Typology and Process in Concept Struc-
turing. Cambridge, MA: The MIT Press, 2000. doi: 10.7551/mitpress/6848.001.0001.
[29] P. H. Thibodeau, R. K. Hendricks, and L. Boroditsky. “How Linguistic Metaphor Scaffolds
Reasoning”. In: Trends in Cognitive Sciences 21.11 (2017), pp. 852–863. doi: 10.1016/j.tics
.2017.07.001.
[30] X. Tong, E. Shutova, and M. Lewis. “Recent Advances in Neural Metaphor Processing:
A Linguistic, Cognitive and Social Perspective”. In: Proceedings of the 2021 Conference of
the North American Chapter of the Association for Computational Linguistics: Human Lan-
guage Technologies. Online: Association for Computational Linguistics, 2021, pp. 4673–
4686. doi: 10.18653/v1/2021.naacl-main.372.
[31] L. Wachowiak and D. Gromann. “Does GPT-3 Grasp Metaphors? Identifying Metaphor
Mappings with Generative Language Models”. In: Proceedings of the 61st Annual Meeting
of the Association for Computational Linguistics (Volume 1: Long Papers). Toronto, Canada:
Association for Computational Linguistics, 2023, pp. 1018–1032. doi: 10.18653/v1/2023
.acl-long.58.
[32] J. Wei, X. Wang, D. Schuurmans, M. Bosma, B. Ichter, F. Xia, E. H. Chi, Q. V. Le, and D.
Zhou. “Chain-of-Thought Prompting Elicits Reasoning in Large Language Models”. In:
Proceedings of the 36th International Conference on Neural Information Processing Systems.
New Orleans, LA: Curran Associates Inc., 2024, pp. 1–14. doi: 10.48550/arXiv.2201.11903.
1116