While corporate and governmental sectors boast numerous terminology databases,
proactive management in the academic realm is scarce, as evidenced by the limited
development of institutional terminology resources [
Recent years have witnessed a notable surge in the adoption of chatbots (CBs) as
conversational agents on websites across various sectors, including education, e-commerce,
healthcare, and entertainment [
This study aims to create a preliminary, rule-based CB focused on enhancing access to information within ATB. The objective is to improve resource accessibility, enabling users to conduct terminological queries and seek guidance on tool usage.
Terminology holds significant importance for communication and organisation in both
public and private institutional settings. Effective terminology management has become
essential, particularly in internationalised contexts, to bolster inter-institutional relations
and enhance intra-institutional communication [
This encompasses tasks such as compiling, describing, documenting, and disseminating
terms, along with resolving inconsistencies and developing terminological tools [
When designing and developing a TB, it is crucial to consider the users' needs [
In the case of the ATB, the "Arqus Questionnaire on Terminology Resources" [
CBs are sophisticated software systems designed to mimic human conversation, employing
Natural Language Processing (NLP) to understand and respond to user inquiries [
In a world that prioritises efficiency and speed, CBs have emerged as a tool capable of
delivering quicker and more effective responses than traditional manual systems [
In the realm of higher education, CBs have the potential to enhance interpersonal
communication, learning experiences, and the delivery of diverse information and
knowledge, given their interactive and intuitive nature [
However, their utilisation in the domain of terminology management, particularly in connection with TB usage, remains constrained. For the purpose of this preliminary work, the rule-based architecture for the ArqusTermBot focuses on enhancing access to information within ATB. To cater to end-user profiles who lack specific knowledge of terminology and the utilisation of TBs, the final objective is to enable users to conduct terminological queries and seek guidance on tool usage.
This study is founded on the fundamental premise that the incorporation of CBs enhances the end-user experience with institutional TBs. The objective of this preliminary research is to enhance ATB by creating a dedicated CB prototype to streamline information access. The subsequent sections elaborate on the key attributes of the multilingual ATB, the utilisation of the Landbot.io platform for CB design and development, and the various stages involved in the creation of ArqusTermBot.
Arqus Termbase is the centralised database of the Arqus European University Alliance
( ), a multilingual network of European academic institutions. This
resource, tailored to the needs of the Alliance [
ATB (Figure 1) presently incorporates seven official languages from partner universities,
and access to terminology is provided from term to concept (semasiological access) and
from concept to term (onomasiological access). It features a simple and advanced
approximate search system [
In designing ATB, primary consideration was given to the diverse users within the
university communities of Arqus Alliance members. They include administrative staff,
teaching and research personnel, and students. Nevertheless, being an open resource, it also
targets external users, including translators, interpreters, interested institutions, and the
general public. Consequently, user-friendly help sections must be developed to cater to user
profiles lacking specific knowledge of terminology and the utilisation of normative
resources.
The CB prototype was constructed using Landbot.io ( ) [
The "Build" section constitutes the central component of the CB design, offering a platform to construct and configure the logic of conversations. This section delineates dialogue flows, actions based on user interactions, and the format of questions and answers. A distinction is made between open-ended questions, allowing the use of natural language (e.g. the user's profile and the institution to which they belong), and multiple-choice questions, where users simply click on the desired topic (e.g. information on the structure of ATB, the languages it covers, the type of terminology it contains, and how it can be accessed).
The "Design" section concentrates on the visual aesthetics of the CB, offering tools to customise its interface and align it with the visual identity of the intended product. Customisation options encompass font type and size, background design, logo selection, and personalisation of the virtual assistant's messages and buttons.
Within the "Settings" section, various options enable the comprehensive configuration of the virtual assistant, ranging from the CB's name to various system customisation parameters. This includes the ability to modify default messages in text insertion segments, help buttons, question fields, or error messages. Additionally, features such as activating the type emulator and setting the average human reading speed are available to enhance the user experience, making it as close as possible to interacting with a real person.
The "Share" section offers various options for sharing and previewing the designed CB. Users can choose how the virtual assistant is displayed, such as a pop-up conversation on the right-hand side of the screen or as a full web page. Moreover, the option to share the CB via a URL enables testing the designed flow without making it publicly accessible to all users.
Lastly, the "Analyze" section focuses on monitoring and evaluating the tool's performance. It offers metrics on user interaction and a comprehensive analysis of the flow, including percentages indicating the path taken by users. Furthermore, user responses are categorised based on the predefined variants, simplifying data collection. This allows CB creators to iteratively optimise the tool, enhancing its effectiveness and aligning it with observed needs from the analysis.
The design of the CB is crucial in establishing the groundwork to meet users' expectations,
aiming to address their needs quickly and effortlessly compared to alternative solutions
[
The initial phase involved comprehensive planning, wherein the objectives of the virtual assistant were delineated, users' needs identified, and the purpose of interactions established. Additionally, the personality of the CB was defined based on the anticipated user interactions. Subsequently, the conversational flow was meticulously designed. A set of questions and their corresponding answers were developed to address the most frequently asked questions anticipated from users. These questions were organised into categories, and a flow diagram was created using the Landbot.io platform. The diagram visually represented various bifurcations based on user responses, outlining the rules that define the CB's architecture. Additionally, specific variables were established to collect userprovided information, while maintaining data privacy. During this phase, a significant challenge involved the difficulty in anticipating and covering the full range of user queries, which can vary widely in scope and specificity. The inherent constraints of a rule-based architecture limit the CB's response flexibility, confining interactions to a pre-established set of conditions and pathways.
The iterative refinement phase involved pilot tests to validate functionality and identify errors. Issues related to level structure and conversation flow were addressed, with options allowing users to return to higher levels without re-entering data. Decision points posing a risk of user dropout were simplified, and closed answers were prioritised over open-ended questions to avoid overwhelming users with excessive natural language. Finally, the visual identity of the CB was crafted to enhance the user experience, drawing inspiration from the existing ATB. The primary challenge of this phase was enhancing the CB's structural fluidity to prevent user frustration and potential abandonment. However, a limitation faced is the inability to precisely ascertain the dropout risk, as this CB prototype has not yet been launched publicly, thereby constraining the ability to analyse and address user behaviour patterns effectively before its release.
To fulfil the requirements associated with information retrieval in ATB, this preliminary study has chosen to develop a rule-based CB prototype. This architectural approach facilitates logical and efficient responses to the queries posed by ATB users, aligning well with the nature of their questions. Consequently, ArqusTermBot is crafted as a virtual assistant with a friendly and empathetic personality, encouraging users to stay engaged until their inquiries are resolved.
The conversational flow is structured into two primary rule groups. The first addresses queries regarding the utilisation of the terminology tool, and resolves doubts about ATB’s structure, terminology access, term typology, languages, entry fields and tool managers. The second elucidates fundamental concepts in the field of terminology for different user profiles, who can seek answers related to the definition of key concepts such as "terminology," "term" and "terminology entry," as well as understand the relationships between them (Figure 2). The goal is to provide users of ATB with the opportunity to acquire a basic understanding of terminology before actively using the resource, ensuring a practical and accurate application of the terminology tool.
Regarding the visual identity of the CB, as depicted in Figure 3, a design aligned with the ATB has been selected. Consequently, the name and visual image incorporate the Alliance logo and the corporate colour palette.
Concerning data collection, the CB is equipped to record data from users interacting with the tool. Variables set during the development phase enable efficient data collection. For instance, the CB can capture information about the user type (teaching and research staff, administrative staff, students, translators, or other) making queries, and the institution to which they belong (Figure 2). However, to uphold confidentiality, the interface refrains from collecting personal data. Instead, it indicates user registration through an internal identifier. In instances of specific inquiries or improvement suggestions, the CB redirects users to the contact section of ATB.
As a result, the incorporation of the ArqusTermBot within ATB will significantly enhance user engagement and information accessibility, especially for end-users with no previous experience using multilingual terminology resources. By streamlining the querying process, the CB reduces the time and effort required for users to handle the TB in order to perform relevant and appropriate searches. Moreover, the CB's ability to provide immediate, accurate responses fosters an interactive and engaging way to navigate terminology content. Therefore, ArqusTermBot represents a significant step towards a more accessible and efficient ATB, not only for the academic community but for other stakeholders.
Notwithstanding, further developments and improvements are considered for this prototype: (1) expanding the rules governing the CB to cover a broader spectrum of questions that users may pose; (2) including the languages of Arqus' member institutions to have a multilingual CB; (3) investigating the development of a neural network based version, which could address several limitations inherent to rule-based CBs, and (4) incorporating metrics analysis to conduct a more comprehensive evaluation of user interaction, which will involve tracking flow analytics, response accuracy, and user satisfaction levels to identify patterns and areas for improvement.
The design of a rule-based CB prototype for ATB serves a dual purpose. Firstly, it acts as an intuitive tool, aiding users in becoming more familiar with this institutional terminology resource. Secondly, it facilitates access to terminological knowledge by addressing the terminology-related inquiries of users interacting with ATB. The benefits of this research extend to the Arqus European University Alliance partner community and anyone interested in applying these technological capabilities to the context of terminology management and TBs.
However, to address a broader spectrum of end-users, this prototype requires further development to overcome issues such as ArqusTermBot’s inability to handle unexpected user inputs and confinement to predefined rules. This will enhance the CB's ability to understand and respond dynamically to complex queries, thereby improving adaptability and response accuracy.
This work was supported by the Spanish Ministry of Universities [Predoctoral Grants for the Training of University Lectures (FPU), FPU21/01204 and FPU20/03089], the European Commission [Arqus European University Alliance, 612247-EPP-1-2019-1-ES-EPPKA2-EURUNIV], the European Commission Erasmus + European Universities [Arqus II, ERASMUSEDU-2022-EURUNIV-1], and the Spanish Ministry of Science and Innovation [Integración transversal de la cultura en una base de conocimiento terminológico medioambiental, PID2020-118369GBI00].