This article considers natural language processing research as a language-ideological practice, looking specifically at the task of toxic language detection, which impacts nearly everybody online through automated content moderation. Industry discourse constructs the category of toxicity through a series of oppositions between civil/healthy/referential/rational and unhealthy/toxic/indexical/emotional. Examples from a toxicity correction dataset demonstrate how this ideology can become encoded algorithmically: a focus on preserving referential content in text “detoxification” results in neglect of important poetic, expressive, and social-indexical functions. Overall, the discursive (...) framings of toxicity construct the ideal speaker in terms of what I call a “referentialist” language ideology, which values rational debate in the (regulated) liberal-democratic public sphere. Ultimately, toxicity detection and other metapragmatic tasks do not merely model the existing pragmatic categories but actively construct them. Toxicity in particular potentially reinforces exclusionary norms of white maleness and promotes online subjectivities that are useful (profitable) to the commercial platforms that shaped the task. Since there is no avoiding NLP as language-ideological practice, independent NLP researchers must acknowledge the political potency of their work by continually reflecting on the categories they work with in relation to models of social-political formation. (shrink)

This survey presents a comprehensive overview of Legal Natural Language Processing (NLP) in the Indian context, with a focus on linguistic diversity across Indian languages and challenges related to equitable access to legal resources. Based on a decade-long analysis of law-centric literature, we trace the evolution of legal NLP research in India, highlighting the adoption of deep learning architectures, pre-trained language models, and domain-specific embeddings. We identify key application areas, such as legal named entity recognition, judgment prediction, legal question answering, (...) case summarization and other key legal NLP tasks. The survey also emphasizes the need for open data sets and reproducible code to foster transparency and scalability in legal NLP research. This work aims to inform researchers, practitioners, and policymakers about current developments, challenges, and future directions in this rapidly evolving field. (shrink)

Emotions are a central aspect of communication. Consequently, emotion analysis (EA) is a rapidly growing field in natural language processing (NLP). However, there is no consensus on scope, direction, or methods. In this paper, we conduct a thorough review of 154 relevant NLP publications from the last decade. Based on this review, we address four different questions: (1) How are EA tasks defined in NLP? (2) What are the most prominent emotion frameworks and which emotions are modeled? (3) Is the (...) subjectivity of emotions considered in terms of demographics and cultural factors? and (4) What are the primary NLP applications for EA? We take stock of trends in EA and tasks, emotion frameworks used, existing datasets, methods, and applications. We then discuss four lacunae: (1) the absence of demographic and cultural aspects does not account for the variation in how emotions are perceived, but instead assumes they are universally experienced in the same manner; (2) the poor fit of emotion categories from the two main emotion theories to the task; (3) the lack of standardized EA terminology hinders gap identification, comparison, and future goals; and (4) the absence of interdisciplinary research isolates EA from insights in other fields. Our work will enable more focused research into EA and a more holistic approach to modeling emotions in NLP. (shrink)

Authorship Attribution (AA) approaches in Natural Language Processing (NLP) are important in various domains, including forensic analysis and cybercrime. However, they pose Ethical, Legal, and Societal Implications/Aspects (ELSI/ELSA) challenges that remain underexplored. Inspired by foundational AI ethics guidelines and frameworks, this research introduces a comprehensive framework of responsible guidelines that focuses on AA tasks in NLP, which are tailored to different stakeholders and development phases. These guidelines are structured around four core principles: privacy and data protection, fairness and non-discrimination, transparency (...) and explainability, and societal impact. Furthermore, to illustrate a practical application of our guidelines, we apply them to a recent AA study that targets identifying and linking potential human trafficking vendors. We believe the proposed guidelines can assist researchers and practitioners in justifying their decisions, assisting ethical committees in promoting responsible practices, and identifying ethical concerns related to NLP-based AA approaches. Our study aims to contribute to ensuring the responsible development and deployment of AA tools. (shrink)

In this tutorial, the meaning of natural language is analysed along the lines proposed by Gottlob Frege and Richard Montague. In building meaning representations, we assume that the meaning of a complex expression derives from the meanings of its components. Typed logic is a convenient tool to make this process of composition explicit. Typed logic allows for the building of semantic representations for formal languages and fragments of natural language in a compositional way. The tutorial ends with the discussion of (...) an example fragment, implemented in the functional programming language Haskell Haskell Team; Jones.. (shrink)

A method to identify ontology components is presented in this article. The method relies on Natural Language Processing (NLP) techniques to extract concepts and relations among these concepts. This method is applied in the legal field to build an ontology dedicated to information retrieval. Legal texts on which the method is performed are carefully chosen as describing and conceptualizing the legal domain. We suggest that this method can help legal ontology designers and may be used while building ontologies dedicated to (...) other tasks than information retrieval. (shrink)

Thirty-Five Years of Research on Neuro-Linguistic Programming. NLP Research Data Base. State of the Art or Pseudoscientific Decoration? The huge popularity of Neuro-Linguistic Programming (NLP) therapies and training has not been accompanied by knowledge of the empirical underpinnings of the concept. The article presents the concept of NLP in the light of empirical research in the Neuro-Linguistic Programming Research Data Base. From among 315 articles the author selected 63 studies published in journals from the Master Journal List of ISI. Out (...) of 33 studies, 18.2% show results supporting the tenets of NLP, 54.5% - results non-supportive of the NLP tenets and 27.3% brings uncertain results. The qualitative analysis indicates the greater weight of the non-supportive studies and their greater methodological worth against the ones supporting the tenets. Results contradict the claim of an empirical basis of NLP. (shrink)

This article is an account of reflections drawn from a total of four explicitation interviews , with two people. The article has both methodological and substantive purposes. Methodologically, we explain the contribution of Neuro-Linguistic Programming in the elicitation of first person accounts through guided introspection. Aspects of NLP have been used by both Vermersch and Petitmengin-Peugeot as means for exploring people's inner worlds. We further elucidate NLP as a set of tools for researchers, emphasising the distinctions these enable researchers to (...) make within the structure of consciousness. As the nature of NLP's methodological contribution to the field of Psychophenomenology has been little articulated, this represents an original feature of this article. Substantively, we show how the application of these tools has generated insights into the fine experiential detail of what we term `moments of knowing'. First, our data suggest that suspension and Epoche, which manifested themselves as unrecognised, or pre-reflective moments of understanding for the participants, may be part of everyday 'knowing'. Second, consciousness appears to be multi-dimensional. In particular it appears that it may be helpful to distinguish between different dimensions of awareness that may be involved when exploring an inner landscape. Third, we consider the apparently transformative effect of the explicitation interview for one of these participants, which emphasises that the interview is an active exploration. Our findings question established views of transformative learning, which hitherto have regarded `critical reflection' as the central process involved in transformative learning. (shrink)

This study investigates a novel approach to speed reading by removing spaces between letters in English text. Using a combination of NLP-based word segmentation and human reading simulations, we examine whether continuous text impairs comprehension or enhances reading speed. Results from 100 sentences indicate that fluent readers can accurately segment text without spaces, achieving reading speeds comparable to or faster than standard spaced text. The findings suggest that spacing is not strictly necessary for comprehension and that predictive processing allows both (...) machines and humans to parse continuous text efficiently. (shrink)

Technology has an increasing influence over consumer lifestyles and behaviors that contribute to the rising prevalence of chronic disease. An opportunity therefore exists for consumer technology companies to use their outsized influence on behavior to meaningfully contribute to improving public health. This paper examines the role of health in corporate reporting and explores how to improve accountability in reporting. A natural language processing (NLP) algorithm and thematic analysis were used to identify patterns in public proxy statements. Results demonstrate the health (...) topics and initiatives that are already present in corporate reporting in the technology sector, and their relative prevalence—showing that while consumer health impact is not absent among for-profit companies, there is no unified language that would allow for performance evaluation and accountability. By establishing transparency, best practices (notably, reliable and auditable metrics), and market recognition, corporate health impact could become an exciting, opportune intervention point for public health. (shrink)

This article describes and refines an experiential distinction which has been highlighted by neuro-linguistic programming (NLP), perceptual positions. When you are imagining a past or future scene, you may perceive it (usually pre-reflectively) from three different viewpoints or perceptual positions. If you are looking at the world from your own point of view, through your own eyes, you are in the first perceptual position. If you are looking at the scene through another person's eyes, appreciating the other person's point of (...) view, you are in the second position. If you are seeing the world from an outside point of view, as an independent observer, you are in the third position. NLP highlighted the fact that our feelings change dramatically according to the perceptual position we adopt. Through a concrete example, Connirae Andreas shows that this distinction does not only concern visual perceptions, but also auditory and kinaesthetic perceptions. She also shows that our visual, auditory and kinaesthetic perceptions may be split in different perceptual positions at the same time, and that this misalignment may cause difficulties. Learning to 'align' our perceptual positions brings us greater wholeness, enables us to become more integrated. (shrink)

This paper separates conditional parameter estima- tion, which consistently raises test set accuracy on statistical NLP tasks, from conditional model struc- tures, such as the conditional Markov model used for maximum-entropy tagging, which tend to lower accuracy. Error analysis on part-of-speech tagging shows that the actual tagging errors made by the conditionally structured model derive not only from label bias, but also from other ways in which the independence assumptions of the conditional model structure are unsuited to linguistic sequences. The (...) paper presents new word-sense disambiguation and POS tagging experiments, and integrates apparently conflicting reports from other recent work. (shrink)

To prevent ordinary people from being harmed by natural language processing (NLP) technology, finding ways to measure the extent to which a language model is biased (e.g., regarding gender) has become an active area of research. One popular class of NLP bias measures are bias benchmark datasets—collections of test items that are meant to assess a language model’s preference for stereotypical versus non-stereotypical language. In this paper, we argue that such bias benchmarks should be assessed with models from the psychometric (...) framework of item response theory (IRT). Specifically, we tie an introduction to basic IRT concepts and models with a discussion of how they could be relevant to the evaluation, interpretation and improvement of bias benchmark datasets. Regarding evaluation, IRT provides us with methodological tools for assessing the quality of both individual test items (e.g., the extent to which an item can differentiate highly biased from less biased language models) as well as benchmarks as a whole (e.g., the extent to which the benchmark allows us to assess not only severe but also subtle levels of model bias). Through such diagnostic tools, the quality of benchmark datasets could be improved, for example by deleting or reworking poorly performing items. Finally, in regards to interpretation, we argue that IRT models’ estimates for language model bias are conceptually superior to traditional accuracy-based evaluation metrics, as the former take into account more information than just whether or not a language model provided a biased response. (shrink)

In this work we present the TULSI system (so named after Turin University Legal Semantic Interpreter), a system to produce automatic annotations of normative documents through the extraction of modificatory provisions. TULSI relies on a deep syntactic analysis and a shallow semantic interpreter that are illustrated in detail. We report the results of an experimental evaluation of the system and discuss them, also suggesting future directions for further improvement.

Is academic integrity research presented from a positive integrity standpoint? This paper uses Natural Language Processing techniques to explore a data set of 8,507 academic integrity papers published between 1904 and 2019.Two main techniques are used to linguistically examine paper titles: bigram analysis and sentiment analysis. The analysis sees the three main bigrams used in paper titles as being “academic integrity”, “academic dishonesty” and “plagiarism detection”. When only highly cited papers are considered, negative integrity bigrams dominate positive integrity bigrams. For (...) example, the 100 most cited academic integrity papers of all time are three times more likely to have “academic dishonesty” included in their titles than “academic integrity”. Similarly, sentiment analysis sees negative sentiment outperforming positive sentiment in the most cited papers.The history of academic integrity research is seen to place the field at a disadvantage due to negative portrayals of integrity. Despite this, analysis shows that change towards positive integrity is possible. The titles of papers by the ten most prolific academic integrity researchers are found to use positive terminology in more cases that not. This suggests an approach for emerging academic integrity researchers to model themselves after. (shrink)

The growing adoption of Artificial Intelligence (AI) in recruitment is often flaunted as a means to enhance efficiency and reduce human biases, despite lack of evidence. This study investigates whether AI, specifically ChatGPT, mitigates or perpetuates existing gender bias in hiring decisions We propose a dual-stage methodological framework combining natural language processing techniques with prompt engineering. We apply the framework to the analysis of a large-scale corpus of job descriptions across various industries and the assessment of how ChatGPT screens resume (...) in response to these descriptions. Our findings indicate that the presence and type of gender bias in job descriptions is echoed in ChatGPT-fueled resume screening. More specifically, job postings in traditionally male-dominated fields emphasize leadership, technical expertise, and agentic traits, while those in female-dominated fields prioritize interpersonal skills and service orientation. ChatGPT exhibits bias in resume screening, favouring male-coded attributes, with candidate selection influenced by names and attributes in resumes perceived as gendered. These findings underscore the risk of AI reinforcing societal biases rather than eliminating them and suggest recommendations to mitigate them. (shrink)

The cultural element is the minimum unit of a cultural system. The systematic categorizing, organizing, and retrieval of the traditional Chinese cultural elements are essential prerequisites for the realization of effective extracting and rational utilization, as well as the prerequisite for exploiting the contemporary value of the traditional Chinese culture. To build an objective, integrated, and reliable classification method and a system of traditional Chinese cultural elements, this study takes the text of Taiping Imperial Encyclopedia in Northern Song Dynasty as (...) the primary data source. The unsupervised word segmentation methods are used to detect Out-of-Vocabulary, and then the segmentation results by the THULAC tool with and without custom dictionary are compared. The TF-IDF algorithm is applied to extract the keywords of cultural elements and the Ochiia coefficient is introduced to create complex networks of traditional Chinese cultural elements. After analyzing the topological characteristics of the network, the community detection algorithm is used to identify the topics of cultural elements. Finally, a “Means-Ends” two-dimensional orthogonal classification system is established to categorize the topics. The results showed that the degree distribution in the complex network of Chinese traditional cultural elements is a scale-free network with γ = 2.28. The network shows a structure of community and hierarchy features. The top 12 communities have taken up to 91.77% of the scale of the networks. Those 12 topics of the traditional Chinese cultural elements are circularly distributed in the orthogonal system of cultural elements’ categorization. (shrink)

This paper investigates the adoption of Neurolinguistic Programming (NLP) in organizational development (OD) interventions. The study synthesizes findings from a comprehensive literature review to uncover significant discoveries, emerging themes, and patterns in the field. The review highlights the potential benefits of NLP, including improved knowledge management, increased efficiency, and enhanced customer satisfaction. However, several impediments have been identified, including those relating to technology, organizational opposition, and the need for competent people. Several strategies and solutions are proposed to address these issues, (...) including investing in data infrastructure, fostering a positive organizational culture, and refining NLP skills. (shrink)

Integrating Artificial Intelligence (AI) tools in Bangladesh's academic landscape has ignited concerns about potentially eroding students' creative writing and critical thinking abilities. While AI offers efficient and personalized learning, there is a looming risk of students using it as a shortcut to success. Educators and policymakers must emphasize the cultivation of writing skills and critical thinking while guiding students to recognize the limitations of AI. While plagiarism checking is crucial for academic integrity, it often falls short in acknowledging students' originality. (...) Natural Language Processing (NLP) emerges as a promising solution, augmenting plagiarism detection by comprehending context, nuances, and novel expressions. This paper proposed a model based on NLP that can incorporate a robust token identification mechanism into AI tools used by the students. A unique identification token is generated at the time of system login and stored in a publicly accessible resource, acting as a distinct marker for user activity. These identification tokens support a plagiarism detection system that continually reviews and compares activity records. It also enables a thorough examination of user involvement across various academic supplementary tools, such as academic writing, presentation creation, image generation, etc. (shrink)

Consultations are key to gather evidence that informs rulemaking. When analysing the feedback received, it is essential for the regulator to appropriately cluster stakeholders’ opinions, as misclustering may alter the representativeness of the positions, making some of them appear majoritarian when they might not be. The European Commission (EC)’s approach to clustering opinions in consultations lacks a standardized methodology, leading to reduced procedural transparency, while making use of computational tools only sporadically. This paper explores how natural language processing (NLP) technologies (...) may enhance the way opinion clustering is currently conducted by the EC. We examine 830 responses to three legislative proposals (the Artificial Intelligence Act, the Digital Markets Act and the Digital Services Act) using both a lexical and semantic approach. We find that some groups (like small and medium companies) have low similarity across all datasets and methodologies despite being clustered in one opinion group by the EC. The same happens for citizens and consumer associations for the consultation run over the DSA. These results suggest that computational tools actually help reduce misclustering of stakeholders’ opinions and consequently allow greater representativeness of the different positions expressed in consultations. They further suggest that the EC could identify a convergent methodology for all its consultations, where such tools are employed in a consistent and replicable rather than occasionally. Ideally, it should also explain when one methodology is preferred to another. This effort should find its way into the Better Regulation toolbox (EC 2023). Our analysis also paves the way for further research to reach a transparent and consistent methodology for group clustering. (shrink)

From the end of 2018 in China, the Big-data Driven Price Discrimination (BDPD) of online consumption raised public debate on social media. To study the consumers’ attitude about the BDPD, this study constructed a semantic recognition frame to deconstruct the Affection-Behavior-Cognition (ABC) consumer attitude theory using machine learning models inclusive of the Labeled Latent Dirichlet Allocation (LDA), Long Short-Term Memory (LSTM), and Snow Natural Language Processing (NLP), based on social media comments text dataset. Similar to the questionnaires published results, this (...) article verified that 61% of consumers expressed negative sentiment toward BDPD in general. Differently, on a finer scale, this study further measured the negative sentiments that differ significantly among different topics. The measurement results show that the topics “Regular Customers Priced High” (69%) and “Usage Intention” (67%) occupy the top two places of negative sentiment among consumers, and the topic “Precision Marketing” (42%) is at the bottom. Moreover, semantic recognition results that 49% of consumers’ comments involve multiple topics, indicating that consumers have a pretty clear cognition of the complex status of the BDPD. Importantly, this study found some topics that had not been focused on in previous studies, such as more than 8% of consumers calling for government and legal departments to regulate BDPD behavior, which indicates that quite enough consumers are losing confidence in the self-discipline of the platform enterprises. Another interesting result is that consumers who pursue solutions to the BDPD belong to two mutually exclusive groups: government protection and self-protection. The significance of this study is that it reminds the e-commerce platforms to pay attention to the potential harm for consumers’ psychology while bringing additional profits through the BDPD. Otherwise, the negative consumer attitudes may cause damage to brand image, business reputation, and the sustainable development of the platforms themselves. It also provides the government supervision departments an advanced analysis method reference for more effective administration to protect social fairness. (shrink)

With the increasing prevalence of Internet usage, Internet-Delivered Psychological Treatment (IDPT) has become a valuable tool to develop improved treatments of mental disorders. IDPT becomes complicated and labor intensive because of overlapping emotion in mental health. To create a usable learning application for IDPT requires diverse labeled datasets containing an adequate set of linguistic properties to extract word representations and segmentations of emotions. In medical applications, it is challenging to successfully refine such datasets since emotion-aware labeling is time consuming. Other (...) known issues include vocabulary sizes per class, data source, method of creation, and baseline for the human performance level. This paper focuses on the application of personalized mental health interventions using Natural Language Processing (NLP) and attention-based in-depth entropy active learning. The objective of this research is to increase the trainable instances using a semantic clustering mechanism. For this purpose, we propose a method based on synonym expansion by semantic vectors. Semantic vectors based on semantic information derived from the context in which it appears are clustered. The resulting similarity metrics help to select the subset of unlabeled text by using semantic information. The proposed method separates unlabeled text and includes it in the next active learning mechanism cycle. Our method updates model training by using the new training points. The cycle continues until it reaches an optimal solution, and it converts all the unlabeled text into the training set. Our in-depth experimental results show that the synonym expansion semantic vectors help enhance training accuracy while not harming the results. The bidirectional Long Short-Term Memory (LSTM) architecture with an attention mechanism achieved 0.85 Receiver Operating Characteristic (ROC curve) on the blind test set. The learned embedding is then used to visualize the activated word's contribution to each symptom and find the psychiatrist's qualitative agreement. Our method improves the detection rate of depression symptoms from online forum text using the unlabeled forum texts. (shrink)

Many NLP tasks rely on accurately estimating word dependency probabilities P(w1|w2), where the words w1 and w2 have a particular relationship (such as verb-object). Because of the sparseness of counts of such dependencies, smoothing and the ability to use multiple sources of knowledge are important challenges. For example, if the probability P(N |V ) of noun N being the subject of verb V is high, and V takes similar objects to V , and V is synonymous to V , then (...) we want to conclude that P(N |V ) should also be reasonably high—even when those words did not cooccur in the training data. (shrink)

Natural Language Processing (NLP) research on AI Safety and social bias in AI has focused on safety for humans and social bias against human minorities. However, some AI ethicists have argued that the moral significance of nonhuman animals has been ignored in AI research. Therefore, the purpose of this study is to investigate whether there is speciesism, i.e., discrimination against nonhuman animals, in NLP research. First, we explain why nonhuman animals are relevant in NLP research. Next, we survey the findings (...) of existing research on speciesism in NLP researchers, data, and models and further investigate this problem in this study. The findings of this study suggest that speciesism exists within researchers, data, and models, respectively. Specifically, our survey and experiments show that (a) among NLP researchers, even those who study social bias in AI, do not recognize speciesism or speciesist bias; (b) among NLP data, speciesist bias is inherent in the data annotated in the datasets used to evaluate NLP models; (c) OpenAI GPTs, recent NLP models, exhibit speciesist bias by default. Finally, we discuss how we can reduce speciesism in NLP research. (shrink)

In natural language processing (NLP), named entity recognition (NER) and semantic classification are essential tasks. NER is a fundamental task, that identify named entities in text such as people, organizations, and locations. In Legal domain, NER is particularly important due to the variety of named entities that appear in legal documents and are important for legal analysis whereas Semantic classification is the process of giving each sentence in a text a semantic label, such as ”fact,””arguments,” or”judgement”. Both NER and Semantic (...) classi- fication are critical component of many NLP applications such as Knowledge base construction and semantic search. Semantic searching of legal documents is a powerful technique that can be used to quickly find the information that you need. By combining NER and semantic classification, one can identify the key elements of a legal document and search for documents that are related to a particular topic or contain a particular type of information. This paper proposes a novel approach to NER and Semantic classification in legal domain using Long short-term memory (LSTM) model. The proposed approach is evaluated on a corpus of legal text and the result of this study suggest that LSTM models are a promising approach for NER and Semantic classification in legal domain and can be used to improve the performance of semantic searching of legal documents. (shrink)

Natural Language Processing (NLP) umfasst Technologien und Methoden zur maschinellen Erkennung und Verarbeitung natürlicher Sprache. Eine zentrale Disziplin in diesem Zusammenhang ist die Computerlinguistik, die zwischen Informatik und Sprachwissenschaft angesiedelt ist. Die Künstliche Intelligenz spielt eine immer größere Rolle. Zum Einsatz kommt NLP bei Chatbots und virtuellen Assistenten, sowohl bei geschriebener als auch bei gesprochener Sprache.

Natural language processing (NLP) has undergone significant changes in its methods, reflecting advances in computing technology and cognitive research. This article reviews the key stages of the evolution of natural language processing methods. The article touches on the topic of the first NLP systems developed, provides justification for the reasons for the complexity of some processed texts and the possible depth of analysis. In addition, it describes not only NLP methods before and after the GPT revolution, but also current trends (...) and prospects in the field of natural language processing. The article allows us to trace how the idea of natural language text has changed during the development of computer analysis methods, as well as to understand what text is in the mirror of natural language processing, what is really the subject of natural language processing research and what cannot be seen through the eyes of a simple researcher who does not use NLP methods. (shrink)

In the field of NLP, “noise” is a notion with many different meanings. This is also true in fields of linguistics in which the analysis of ecological data is central.In studies involving corpus linguistic methods, noise management is an essential process for data collection, data structuration and data analysis. Paradoxically, this step is almost never developed, or completely ignored.In this paper, we propose to focus on the management of noise during the various stages classically defined in the processing of an (...) oral corpus: data structuration (including transcription), data enrichment (including annotation) and data analysis.We show that noise management is a cross-cutting issue: it needs to be taken into account during the phases of structuration, data enrichment and analysis of linguistic corpora. More specifically, the creation of corpora of children’s productions leads us to manage a particular type of noise: the variation regarding the so-called “standard” forms. For noise management, we show that there is no ideal methodology: noise management is specific to each project. Moreover, decisions taken at one stage necessarily influence the others: a noise generated during data structuration may cause a silence during automatic processing; a noise “cleaned up” during the processing phase may generate a silence during analysis. (shrink)

Despite its centrality in the philosophy of cognitive science, there has been little prior philosophical work engaging with the notion of representation in contemporary NLP practice. This paper attempts to fill that lacuna: drawing on ideas from cognitive science, I introduce a framework for evaluating the representational claims made about components of neural NLP models, proposing three criteria with which to evaluate whether a component of a model represents a property and operationalising these criteria using probing classifiers, a popular analysis (...) technique in NLP (and deep learning more broadly). The project of operationalising a philosophically-informed notion of representation should be of interest to both philosophers of science and NLP practitioners. It affords philosophers a novel testing-ground for claims about the nature of representation, and helps NLPers organise the large literature on probing experiments, suggesting novel avenues for empirical research. (shrink)

Gatekeeping, or determining membership of your group, is crucial to science: the moniker ‘scientific’ is a stamp of epistemic quality or even authority. But gatekeeping in science is fraught with dangers. Gatekeepers must exclude bad science, science fraud and pseudoscience, while including the disagreeing viewpoints on which science thrives. This is a difficult tightrope, not least because gatekeeping is a human matter and can be influenced by biases such as groupthink. After spelling out these general tensions around gatekeeping in science, (...) we shed light on them with a case study from psychology. This concerns whether academic psychologists rightly or wrongly classify the applied-psychology framework of NLP (‘neuro-linguistic programming’) as unscientific and even pseudoscientific. This example of gatekeeping is particularly instructive because both the NLP community and the psychology community, we argue, make legitimate but also illegitimate moves. This case gives rise to several general insights about gatekeeping in science more generally. (shrink)

Knowledge claims are abundant in the literature on large language models (LLMs); but can we say that GPT-4 truly "knows" the Earth is round? To address this question, we review standard definitions of knowledge in epistemology and we formalize interpretations applicable to LLMs. In doing so, we identify inconsistencies and gaps in how current NLP research conceptualizes knowledge with respect to epistemological frameworks. Additionally, we conduct a survey of 100 professional philosophers and computer scientists to compare their preferences in knowledge (...) definitions and their views on whether LLMs can really be said to know. Finally, we suggest evaluation protocols for testing knowledge in accordance to the most relevant definitions. (shrink)

Identifying, classifying, and analyzing arguments in legal discourse has been a prominent area of research since the inception of the argument mining field. However, there has been a major discrepancy between the way natural language processing (NLP) researchers model and annotate arguments in court decisions and the way legal experts understand and analyze legal argumentation. While computational approaches typically simplify arguments into generic premises and claims, arguments in legal research usually exhibit a rich typology that is important for gaining insights (...) into the particular case and applications of law in general. We address this problem and make several substantial contributions to move the field forward. First, we design a new annotation scheme for legal arguments in proceedings of the European Court of Human Rights (ECHR) that is deeply rooted in the theory and practice of legal argumentation research. Second, we compile and annotate a large corpus of 373 court decisions (2.3M tokens and 15k annotated argument spans). Finally, we train an argument mining model that outperforms state-of-the-art models in the legal NLP domain and provide a thorough expert-based evaluation. All datasets and source codes are available under open lincenses athttps://github.com/trusthlt/mining-legal-arguments. (shrink)

This paper addresses the treatment of noise caused by word omissions in a corpus of school writings, in order to facilitate their subsequent automatic processing. While a normalization step may facilitate the processing of these texts, certain linguistic expressions remain challenging to comprehend, particularly in instances where the writer omits words from the text. The present contribution proposes three automatic and semi-automatic potential solutions to this problem. The first method employs a "mask" token in the form of xxx. The second (...) is a semi-automatic approach whereby each morpho-syntactic category proposed during normalization is replaced by the corresponding "prototypical word." The third involves a FlauBERT method, using this language model to "reconstruct" the most probable token in the text. The three methods are evaluated quantitatively, and the results obtained using method 3, which proved to be the most effective in the context of our research, are also presented qualitatively. (shrink)

Responsible artificial intelligence (AI) requires integrating ethical awareness into the full process of designing and developing AI, including ethics-based auditing of AI technology. We claim the Capability Approach (CA) of Sen and Nussbaum grounds AI ethics in essential human freedoms and can increase awareness of the moral dimension in the technical decision making of developers and data scientists constructing data-centric AI systems. Our use of CA focuses awareness on the ethical impact that day-to-day technical decisions have on the freedom of (...) data subjects to make choices and live meaningful lives according to their own values. For internal auditing of AI technology development, we design and develop a light-weight ethical auditing tool (LEAT) that uses simple natural language processing (NLP) techniques to search design and development documents for relevant ethical characterizations. We describe how CA guides our design, demonstrate LEAT on both principle- and capabilities-based use cases, and characterize its limitations. (shrink)

Artificial intelligence (AI) is revolutionizing nursing by enhancing decision‐making, patient monitoring, and efficiency. Machine learning, natural language processing (NLP), and predictive analytics claim to improve safety and automate tasks. However, a structured analysis of AI applications is necessary to ensure their effective implementation in nursing practice. This umbrella review aimed to synthesize existing systematic reviews on AI applications in nursing care, providing a comprehensive analysis of its benefits, challenges, and ethical implications. By consolidating findings from multiple sources, this review seeks (...) to offer evidence‐based insights to guide the effective and responsible integration of AI in nursing practice. A systematic umbrella review approach was employed following PRISMA guidelines. Multiple databases, including PubMed, CINAHL, Scopus, Web of Science, and IEEE Xplore, were searched for review articles published between 2015 and 2024. Findings were synthesized thematically to identify key trends, benefits, limitations, and research gaps. This review synthesized 13 studies, emphasizing AI's impact on clinical decision support, patient monitoring, nursing education, and workflow optimization. AI enhances early disease detection, minimizes diagnostic errors, and automates documentation, improving efficiency. However, data privacy risks, biases, ethical concerns, and limited AI literacy hinder integration. AI presents significant opportunities for improving nursing care, yet its successful implementation requires addressing ethical, legal, and practical challenges. Adequate AI training, robust data governance frameworks, and policies ensuring responsible AI use are essential for its integration into nursing practice. Future research should explore long‐term AI impact, training models for nurses, and strategies to balance AI‐driven efficiency with human‐centered care. (shrink)

As if 2020 were not a peculiar enough year, its fifth month has seen the relatively quiet publication of a preprint describing the most powerful Natural Language Processing (NLP) system to date, GPT-3 (Generative Pre-trained Transformer-3), by Silicon Valley research firm OpenAI. Though the software implementation of GPT-3 is still in its initial Beta release phase, and its full capabilities are still unknown as of the time of this writing, it has been shown that this Artificial Intelligence can comprehend prompts (...) in natural language, on virtually any topic, and generate relevant, original text content that is indistinguishable from human writing. Moreover, access to these capabilities, in a limited yet worrisome enough extent, is available to the general public as of the time of this writing. This paper presents select examples of original content generated by the author using GPT-3. These examples illustrate some of the capabilities of GPT-3 in comprehending prompts in natural language and generating convincing content in response. We use these examples to raise specific, fundamental questions pertaining to the intellectual property of this content and the potential use of GPT-3 to facilitate plagiarism. Our goal is to instigate not just a sense of urgency, but of a present tardiness on the part of the academic community in addressing these questions. (shrink)

This article argues that due to the difficulty in governing AI, it is essential to develop measures implemented early in the AI research process. The goal of dual use considerations is to create robust strategies that uphold AI’s integrity while protecting societal interests. The challenges of applying dual use frameworks to AI research are examined and dual use and dual use research of concern (DURC) are defined while highlighting the difficulties in balancing the technology’s benefits and risks. AI’s dual use (...) potential is discussed, particularly in areas like NLP and LLMs, and the need for early consideration of dual use risks to ensure ethical and secure development is underscored. In the section on shared responsibilities in AI research and avenues for mitigation strategies the importance of early-stage risk assessments and ethical guidelines to mitigate misuse is emphasized, accentuating self-governance within scientific communities and structured measures like checklists and pre-registration to promote responsible research practices. The final section argues that research ethics committees play a crucial role in evaluating the dual use implications of AI technologies within the research pipeline. The need for tailored ethics review processes is articulated, drawing parallels with medical research ethics committees. (shrink)

Knowing that language is often used as a classifier of human intelligence and the development of systems that understand human language remains a challenge all the time (Kryeziu & Shehu, 2022). Natural Language Processing is a very active field of study, where transformers have a key role. Transformers function based on neural networks and they are increasingly showing promising results. One of the first major contributions to transfer learning in Natural Language Processing was the use of pre-trained word embeddings in (...) 2010 (Joseph, Lev, & Yoshua, 2010). Pre-trained models like ELMo (Matthew, et al., 2018) and BERT (Delvin, et al., 2019) are trained on large corpora of unlabeled text and as a result learning from text representations has achieved good performance on many of the underlying tasks on datasets from different domains. Pre-training in the language model has proven that there has been an improvement in some aspects of natural language processing, based on the paper (Dai & Le, 2015). In present paper, we will pre-train BERT on the task of Masked Language Modeling (MLM) with the Albanian language dataset (alb_dataset) that we have created for this purpose (Kryeziu et al., 2022). We will compare two approaches: training of BERT using the available OSCAR dataset and using our alb_dataset that we have collected. The paper shows some discrepancies during training, especially while evaluating the performance of the model. (shrink)

Cet article détaille la méthodologie mise en place dans le projet EMOLEX qui a abouti à la mise à disposition de corpus multilingues, d’interfaces d’interrogation et d’analyse de ces corpus ainsi que d’applications permettant d’exploiter les analyses linguistiques portant sur le lexique des affects dans cinq langues européennes.

This article examines how personification metaphors shape public perceptions of generative AI through a computational analysis of Tweets posted during the emergence of ChatGPT (2022–2023). Using a hybrid methodology combining rule-based NLP and the DeepMet neural network model, we identify AI personification patterns and discuss their ideological implications. Results reveal that most metaphors employ Subject–Verb–Object constructions framing AI as an active agent. The major source domains are COGNITION (e.g. think), ACTION/CHANGE (e.g., replace), COMMUNICATION (e.g., listen), and EMOTION (e.g., love, fear). (...) Sentiment analysis shows predominantly positive attitudes, particularly when framing AI as a collaborative assistant, while negative sentiment is strongly associated with job displacement concerns. These patterns suggest that users assign agency and autonomy to AI systems with critical implications: personification underpins an implicit social contract where augmentation is embraced and replacement resisted, yet it simultaneously obscures accountability by positioning AI as an independent actor rather than a product of corporate design and deployment. Moreover, as LLM outputs become more sophisticated, the figurative distance between computational systems and human behavior risks collapsing entirely. Our findings show that metaphorical language dynamically constructs social realities about generative AI, raising stakes for public trust, responsibility attribution, and technology governance. (shrink)

Artificial intelligence (AI) systems are increasingly being used not only to classify and analyze but also to generate images and text. As recent work on the content produced by text and image Generative AIs has shown (e.g., Cheong et al., 2024, Acerbi & Stubbersfield, 2023), there is a risk that harms of representation and bias, already documented in prior AI and natural language processing (NLP) algorithms may also be present in generative models. These harms relate to protected categories such as (...) gender, race, age, and religion. There are several kinds of harms of representation to consider in this context, including stereotyping, lack of recognition, denigration, under-representation, and many others (Crawford in Soundings 41:45–55, 2009; in: Barocas et al., SIGCIS Conference, 2017). Whereas the bulk of researchers’ attention thus far has been given to stereotyping and denigration, in this study we examine ‘exnomination’, as conceived by Roland Barthes (1972), of religious groups. Our case study is DALL-E, a tool that generates images from natural language prompts. Using DALL-E mini, we generate images from generic prompts such as “religious person.” We then examine whether the generated images are recognizably members of a nominated group. Thus, we assess whether the generated images normalize some religions while neglecting others. We hypothesize that Christianity will be recognizably represented more frequently than other religious groups. Our results partially support this hypothesis but introduce further complexities, which we then explore. (shrink)

Argumentation mining is an application of natural language processing (NLP) that emerged a few years ago and has recently enjoyed considerable popularity, as demonstrated by a series of international workshops and by a rising number of publications at the major conferences and journals of the field. Its goals are to identify argumentation in text or dialogue; to construct representations of the constellation of claims, supporting and attacking moves (in different levels of detail); and to characterize the patterns of reasoning that (...) appear to license the argumentation. Furthermore, recent work also addresses the difficult tasks of evaluating the persuasiveness and quality of arguments. Some of the linguistic genres that are being studied include legal text, student essays, political discourse and debate, newspaper editorials, scientific writing, and others. The book starts with a discussion of the linguistic perspective, characteristics of argumentative language, and their relationship to certain other notions such as subjectivity. Besides the connection to linguistics, argumentation has for a long time been a topic in Artificial Intelligence, where the focus is on devising adequate representations and reasoning formalisms that capture the properties of argumentative exchange. It is generally very difficult to connect the two realms of reasoning and text analysis, but we are convinced that it should be attempted in the long term, and therefore we also touch upon some fundamentals of reasoning approaches. Then the book turns to its focus, the computational side of mining argumentation in text. We first introduce a number of annotated corpora that have been used in the research. From the NLP perspective, argumentation mining shares subtasks with research fields such as subjectivity and sentiment analysis, semantic relation extraction, and discourse parsing. Therefore, many technical approaches are being borrowed from those (and other) fields. We break argumentation mining into a series of subtasks, starting with the preparatory steps of classifying text as argumentative (or not) and segmenting it into elementary units. Then, central steps are the automatic identification of claims, and finding statements that support or oppose the claim. For certain applications, it is also of interest to compute a full structure of an argumentative constellation of statements. Next, we discuss a few steps that try to 'dig deeper': to infer the underlying reasoning pattern for a textual argument, to reconstruct unstated premises (so-called 'enthymemes'), and to evaluate the quality of the argumentation. We also take a brief look at 'the other side' of mining, i.e., the generation or synthesis of argumentative text. The book finishes with a summary of the argumentation mining tasks, a sketch of potential applications, and a—necessarily subjective—outlook for the field. Table of Contents: Preface / Acknowledgments / Introduction / Argumentative Language / Modeling Arguments / Corpus Annotation / Finding Claims / Finding Supporting and Objecting Statements / Deriving the Structure of Argumentation / Assessing Argumentation / Generating Argumentative Text / Summary and Perspectives / Bibliography / Authors' Biographies / Index. (shrink)

Named entity recognition (NER) is a very relevant task for text information retrieval in natural language processing (NLP) problems. Most recent state-of-the-art NER methods require humans to annotate and provide useful data for model training. However, using human power to identify, circumscribe and label entities manually can be very expensive in terms of time, money, and effort. This paper investigates the use of prompt-based language models (OpenAI’s GPT-3) and weak supervision in the legal domain. We apply both strategies as alternative (...) approaches to the traditional human-based annotation method, relying on computer power instead human effort for labeling, and subsequently compare model performance between computer and human-generated data. We also introduce combinations of all three mentioned methods (prompt-based, weak supervision, and human annotation), aiming to find ways to maintain high model efficiency and low annotation costs. We showed that, despite human labeling still maintaining better overall performance results, the alternative strategies and their combinations presented themselves as valid options, displaying positive results and similar model scores at lower costs. Final results demonstrate preservation of human-trained models scores averaging 74.0% for GPT-3, 95.6% for weak supervision, 90.7% for GPT + weak supervision combination, and 83.9% for GPT + 30% human-labeling combination. (shrink)

Due to huge advancements in natural language processing (NLP) and machine learning, chatbots are gaining significance in the field of customer service. For users, it may be hard to distinguish whether they are communicating with a human or a chatbot. This brings ethical issues, as users have the right to know who or what they are interacting with (European Commission in Regulatory framework proposal on artificial intelligence. https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai, 2022). One of the solutions is to include a disclosure at the start (...) of the interaction (e.g., “this is a chatbot”). However, companies are reluctant to use disclosures, as consumers may perceive artificial agents as less knowledgeable and empathetic than their human counterparts (Luo et al. in Market Sci 38(6):937–947, 2019). The current mixed methods study, combining qualitative interviews (n = 8) and a quantitative experiment (n = 194), delves into users’ responses to a disclosed vs. undisclosed customer service chatbot, focusing on source orientation, anthropomorphism, and social presence. The qualitative interviews reveal that it is the willingness to help the customer and the friendly tone of voice that matters to the users, regardless of the artificial status of the customer care representative. The experiment did not show significant effects of the disclosure (vs. non-disclosure). Implications for research, legislators and businesses are discussed. (shrink)