Search

Our research strives to examine how simulation models of logistics systems can be produced automatically from verbal descriptions in natural language and how human experts and artificial intelligence (AI)-based systems can collaborate in the domain of simulation modelling. We demonstrate that a framework constructed upon the refined GPT-3 Codex is capable of generating functionally valid simulations for queuing and inventory management systems when provided with a verbal explanation. As a result, the language model could produce simulation models for inventory and process control. These results, along with the rapid improvement of language models, enable a significant simplification of simulation model development. Our study offers guidelines and a design of a natural language processing-based framework on how to build simulation models of logistics systems automatically, given the verbal description. In generalised terms, our work offers a technological underpinning of human-AI collaboration for the development of simulation models. [ABSTRACT FROM AUTHOR]

ABSTRACT Chatbots have emerged as ubiquitous tools for enhancing user interaction across various platforms, from customer service to personal assistance. They are computer programs that simulate and process human conversation, either written, spoken or both. However, developing efficient chatbots remains a challenge, primarily due to the intricate nature of critical components of chatbots like natural language understanding (NLU) requiring a subscription from intent recognition providers like Dialogflow and Amazon Lex. This makes chatbots closely linked to NLP services and can be locked in. Recently, various research studies have provided solutions to reduce the workload of developers and designers. These approaches have proposed model‐driven development via domain‐specific languages (DSLs), which make the chatbot development process more accessible and more automated. This advancement aims to enhance effectiveness in chatbot development by leveraging DSLs. This study aims to provide a comprehensive overview of DSLs for developing chatbots, with the first contribution comprising various research topics, tools, approaches, and technologies employed to implement DSLs. Second, this work aims to assess and contrast the primary DSLs currently available for chatbot development, focusing on presenting the key elements used in constructing these DSLs. Third, this study identifies the challenges and limitations of using DSLs in chatbot development. [ABSTRACT FROM AUTHOR]

An enigma for human languages is that children learn to understand words in their mother tongue extremely fast. The cognitive sciences have not been able to fully understand the mechanisms behind this highly efficient learning process. In order to provide at least a partial answer to this problem, I have developed a cognitive model of the semantics of natural language in terms of conceptual spaces. I present a background to conceptual spaces and provide a brief summary of their main features, in particular how it handles learning of concepts. I then apply the model to give a geometric account of the semantics of different word classes. In particular, I propose a “single‐domain hypotheses” for the semantics of all word classes except nouns. These hypotheses provide a partial answer to the enigma of how words are learned. Next, a dynamic cognitive model of events is introduced that replaces and extends the function of thematic roles. I apply it to analyze the meanings of different kinds of verbs. I argue that the model also explains some aspects of syntactic structure. In particular, I propose that a sentence typically refers to an event. Some further applications of conceptual spaces are briefly presented. [ABSTRACT FROM AUTHOR]

ABSTRACT Artificial intelligence (AI) technology has brought groundbreaking changes to the healthcare domain. Specifically, the integration of a medical dialogue system (MDS) has facilitated interactions with patients, identifying meaningful information such as symptoms and medications from their dialogue history to generate appropriate responses. However, shortcomings arise when MDS lacks access to the patient's cumulative history or prior domain knowledge, resulting in the generation of inaccurate responses. To address this challenge, we propose a medical domain‐incremental task‐oriented dialogue generator using experience replay (MeDi‐TODER) that applies the continual learning technique to the medical task‐oriented dialogue generator. By strategically sampling and storing exemplars from previous domains and rehearsing it as it learns, the model effectively retains knowledge and can respond to the novel domains. Extensive experiments demonstrated that MeDi‐TODER significantly outperforms other models that lack continual learning in both natural language generation and natural language understanding. Notably, our proposed method achieves the highest scores, surpassing the upper‐class benchmarks. [ABSTRACT FROM AUTHOR]

Different people listening to the same story may converge upon a largely shared interpretation while still developing idiosyncratic experiences atop that shared foundation. What linguistic properties support this individualized experience of natural language? Here, we investigate how the "concrete--abstract" axis--the extent to which a word is grounded in sensory experience--relates to within- and across-subject variability in the neural representations of language. Leveraging a dataset of human participants of both sexes who each listened to four auditory stories while undergoing functional magnetic resonance imaging, we demonstrate that neural representations of "concreteness" are both reliable across stories and relatively unique to individuals, while neural representations of "abstractness" are variable both within individuals and across the population. Using natural language processing tools, we show that concrete words exhibit similar neural representations despite spanning larger distances within a high-dimensional semantic space, which potentially reflects an underlying representational signature of sensory experience--namely, imageability-- shared by concrete words but absent from abstract words. Our findings situate the concrete--abstract axis as a core dimension that supports both shared and individualized representations of natural language. [ABSTRACT FROM AUTHOR]

This paper introduces a comprehensive framework for the detection and identification of malicious smart contracts, emphasizing their vulnerabilities. The framework leverages the capabilities of GPT‐3, which have been adapted and fine‐tuned for binary and multi‐class classification tasks. To the best of our knowledge, this study is the first to explore the use of GPT‐3 specifically for detecting and identifying malicious smart contracts. The framework addresses previously unexplored research questions and provides insightful answers through rigorous experimentation. The contributions of this work include proposing a novel approach, pioneering the adaptation of GPT‐3 for this purpose, and offering valuable insights into the detection of malicious smart contracts and vulnerabilities. Notably, our research reveals that GPT‐3 excels not only in understanding natural language but also in decoding the secrets embedded in numerical codes like opcodes. This finding extends the applicability of GPT‐3 beyond language‐based tasks and highlights its potential in enhancing smart contract security. [ABSTRACT FROM AUTHOR]

Referring expression comprehension (REC) system solves a task to localize objects in a given image, based on natural language expression. We propose a novel approach to adapting the pre-trained REC model for the manufacturing domain. Despite significant advances in REC research, current REC datasets fail to recognize objects from specific yet important domains such as manufacturing due to the absence of domain-specific samples during training. Thus, we introduce a synthetic data-based domain adaptation approach for REC. To adapt a REC model to the manufacturing domain, we generated a synthetic REC dataset RefMD that consists of two sub-datasets: (1) dataset for manufacturing object classification, and (2) dataset for REC adaptation to manufacturing. Each dataset serves as one step toward the REC adaptation. Adaptation of the object classification network (visual backbone) is carried out by training ResNet50 on domain-specific labeled data, while the REC adaptation completes with the adaptation of modules in RealGIN altogether. The manufacturing domain-adapted model is further enhanced with the capability to handle ambiguous referring expressions through human-in-the-loop (HITL) interaction. The experiments on 3D-printed manufacturing objects demonstrate that the interactive REC model can accurately comprehend human instructions with an 82% accuracy. This paper introduces an approach to facilitate domain adaptation using solely synthetic data for the specific case of the manufacturing domain. However, the proposed adaptation methodology can be applied to any other domain by following the same synthetic data generation process. [ABSTRACT FROM AUTHOR]

The mapping process generally involves intricate operations, such as symbol design, layout design, and text annotation, demanding a high level of professional expertise. The high requirement for map producers hinders the promotion and widespread adoption of mapping. Consequently, researchers are concentrating on techniques to automate and enhance the intelligence of the mapping process. For example, some studies attempt to train deep learning models for mapping, including methods like map style transfer. However, these approaches typically treat the entire map as a global input and generate a new map as output, lacking the flexibility to consider and control detailed elements within a map. Therefore, in this paper, we propose a large language model-based intelligent mapping framework, termed MapGPT, which can be used for mapping by considering the map as an integration of various map elements. Specifically, multiple professional mapping tools are designed in MapGPT, and each tool is designed to control a corresponding map element. With these tools, a large language model is used to first understand the demand of users based on mere natural language descriptions, and subsequently automatically invoke appropriate tools in sequence to generate a map. Furthermore, by utilizing a memory component to store interaction information, users can interact with MapGPT through conversation to adjust map elements such as color and position. In conclusion, MapGPT offers user-friendly mapping experience, showing potential to be a mapping assistant for professional map producers. A comprehensive demonstration of this framework is provided in a visual case study video, accessible at . [ABSTRACT FROM AUTHOR]

This paper presents a procedure for and evaluation of using a semantic similarity metric as a loss function for neural source code summarization. Code summarization is the task of writing natural language descriptions of source code. Neural code summarization refers to automated techniques for generating these descriptions using neural networks. Almost all current approaches involve neural networks as either standalone models or as part of a pretrained large language models, for example, GPT, Codex, and LLaMA. Yet almost all also use a categorical cross‐entropy (CCE) loss function for network optimization. Two problems with CCE are that (1) it computes loss over each word prediction one‐at‐a‐time, rather than evaluating a whole sentence, and (2) it requires a perfect prediction, leaving no room for partial credit for synonyms. In this paper, we extend our previous work on semantic similarity metrics to show a procedure for using semantic similarity as a loss function to alleviate this problem, and we evaluate this procedure in several settings in both metrics‐driven and human studies. In essence, we propose to use a semantic similarity metric to calculate loss over the whole output sentence prediction per training batch, rather than just loss for each word. We also propose to combine our loss with CCE for each word, which streamlines the training process compared to baselines. We evaluate our approach over several baselines and report improvement in the vast majority of conditions. [ABSTRACT FROM AUTHOR]

Making sense of natural language and narratives requires building and manipulating a situation model by adding incoming information to the model and using the context stored in the model to comprehend subsequent details and events. Situation model maintenance is supported by the default mode network (DMN), but comprehension of the individual moments in the narrative relies on access to the conceptual store within the semantic system. The present study examined how these systems are engaged by different narrative content to investigate whether highly informative, or semantic, content is a particularly strong driver of semantic system activation compared with contextually driven content that requires using the situation model, which might instead engage DMN regions. The study further investigated which subregions of the graded semantic hub in the left anterior temporal lobe (ATL) were engaged by the type of narrative content. To do this, we quantified the semantic, pragmatic, social, ambiguous, and emotional content for each sentence in a complete narrative, the English translation of The Little Prince. Increased activation in the transmodal hub in the ventral ATL was only observed for high semantic (i.e., informative) relative to low semantic sentences. Activation in the dorsolateral and ventrolateral ATL subregions was observed for both high relative to low semantic and social content sentences, but the ventrolateral ATL effects were more extensive in the social condition. There was high correspondence between the social and pragmatic content results, particularly in the ventrolateral ATL. We argue that the ventrolateral ATL may be particularly engaged by internal, or endogenous, processing demands, aided by functional connections between the anterior middle temporal gyrus and the DMN. Pragmatic and social content may have driven endogenous processing given the pervasive and plot-progressing nature of this content in the narrative. We put forward a revised account of how the semantic system is engaged in naturalistic contexts, a critical step toward better understanding real-world semantic and social processing. [ABSTRACT FROM AUTHOR]

Business Intelligence is the key technologies that ensures effective decision making through extracting relevant information and providing adapted systems as the Data Warehouses. To access decisional information, the decision maker should express his requirements in Natural Language interfaces without any technical skills, avoiding the IT-Designer intervention. Often, the decision maker's requirements are expressed as WH-questions (What, Who, Where, etc.) or Keyword-like questions. In this paper, we emphasize on a Why-Question asked in Business Intelligence context. This question has not been well dealt in the literature in terms of produced answers. Indeed, to respond this type of question, it is necessary to provide explanations. These explanations are determined by identifying causal relationships between the phenomenon highlighted in the Why-Question and factors that can influence this phenomenon. In this context, we propose an approach on which a system can address a causality problem related to answering a decisional Why-Question. To validate our approach a tool called BI Why Q/A is developed. In order evaluate our proposal in terms of efficiency and relevance, a set of experimental studies is carried out and presented. [ABSTRACT FROM AUTHOR]

Depression is a common pregnancy complication yet is often under-detected and, subsequently, undertreated. Data collected through mobile health tools may be used to support the identification of depression symptoms in pregnancy. An observational cohort study of 2062 pregnancies collected self-reports of patient history, mood, pregnancy-specific symptoms, and written language using a prenatal support app. These app inputs were used to model depression risk in subsequent 30- and 60-day periods throughout pregnancy. A selective inference lasso modeling approach examined the individual and additive value of each type of patient-reported app input. Depression models ranged in predictive power (AUC value of 0.64–0.83), depending on the type of inputs. The most predictive model included personal history, daily mood, and acute pregnancy-related symptoms (e.g., severe vomiting, cramping). Across models, daily mood was the strongest indicator of depression symptoms in the following month. Models that retained natural language inputs typically improved predictive accuracy and offered insight into the lived context associated with experiencing depression. Our findings are not generalizable beyond a digitally literate patient population that is self-motivated to report data during pregnancy. Simple patient reported data, including sparse language, shared directly via digital tools may support earlier depression symptom identification and a more nuanced understanding of depression context. [Display omitted] • Self-reported pregnancy app data can provide insight into depression symptoms. • Mood is a strong indicator of moderate-to-severe depression symptoms. • The best depression models include patient history and acute pregnancy symptoms. • Natural (written) language provides context to the drivers of mood and depression. [ABSTRACT FROM AUTHOR]

The detection of essays written by AI compared to those authored by students is increasingly becoming a significant issue in educational settings. This research examines various numerical text representation techniques to improve the classification of these essays. Utilizing a diverse dataset, we undertook several preprocessing steps, including data cleaning, tokenization, and lemmatization. Our system analyzes different text representation methods such as Bag of Words, TF-IDF, and fastText embeddings in conjunction with multiple classifiers. Our experiments showed that TF-IDF weights paired with logistic regression reached the highest accuracy of 99.82%. Methods like Bag of Words, TF-IDF, and fastText embeddings achieved accuracies exceeding 96.50% across all tested classifiers. Sentence embeddings, including MiniLM and distilBERT, yielded accuracies from 93.78% to 96.63%, indicating room for further refinement. Conversely, pre-trained fastText embeddings showed reduced performance, with a lowest accuracy of 89.88% in logistic regression. Remarkably, the XGBoost classifier delivered the highest minimum accuracy of 96.24%. Specificity and precision were above 99% for most methods, showcasing high capability in differentiating between student-created and AI-generated texts. This study underscores the vital role of choosing dataset-specific text representations to boost classification accuracy. [ABSTRACT FROM AUTHOR]

How can modern neural networks like language models be useful to the field of language acquisition, and more broadly cognitive science, if they are not a priori designed to be cognitive models? As developments towards natural language understanding and generation have improved leaps and bounds, with models like GPT‐4, the question of how they can inform our understanding of human language acquisition has re‐emerged. As such, it is critical to examine how in practice linking hypotheses between models and human learners can be safely established. To address these questions, we propose a model taxonomy, including four modelling approaches, each having differing goals, from exploratory hypothesis generation to hypothesis differentiation and testing. We show how the goals of these approaches align with the overarching goals of science and linguistics by connecting our taxonomy to the realist versus instrumentalist approaches in philosophy of science. We survey recent work having adopted each of our modelling approaches and address the importance of computational modelling in language acquisition studies. [ABSTRACT FROM AUTHOR]

One of the foundational subjects in both artificial intelligence (AI) and database technologies is natural language interfaces for databases (NLIDB). The primary goal of NLIDB is to enable users to interact with databases using natural languages such as English, Arabic, and French. While many existing NLIDBs rely on linguistic operations to meet the challenges of user's ambiguity existing in natural language queries (NLQ), there is currently a growing emphasis on utilizing inductive logic programming (ILP) to develop natural language processing (NLP) applications. This is because ILP reduces the requirement for linguistic expertise in building NLP systems. This paper outlines a methodology for automating the construction of NLIDB. This method utilizes ILP to derive transfer rules that directly translate NLQ into a clear and unambiguous logical query, which subsequently translatable into database query languages (DQL). To acquire these rules, our system was trained within a corpus consisting of parallel examples of NLQs and their logical interpretations. The experimental results demonstrate the promise of this approach, as it enables the direct translation of all NLQs with grammatical structures similar to those already present in the trained corpus into a logical query. [ABSTRACT FROM AUTHOR]

The traditional English grammar error correction system has problems such as poor error recognition precision and error correction success rate to be improved. Therefore, in this study, data augmentation techniques were used to transform and process error correcting English texts, and a rule-based and shallow neural network-based English text grammar correction model was constructed. The experimental results showed that the grammar error generation (GEG), rule-based (RB), classification-based (CB), and recurrent neural network (RNN) models achieved accuracy rates of 93.92%, 82.17%, 79.41%, and 88.09% in correcting grammar errors on 1702641 test sentences in the One Billion word corpus, respectively. The experimental results showed that the English grammar error correction model designed in this study had a strong error correction ability, but the computational efficiency was low. The research results significantly improved the accuracy and generalization ability of English grammar correction, optimized learning costs, and brought positive impacts to educational applications, providing strong support for the development of intelligent English grammar correction. [ABSTRACT FROM AUTHOR]

This paper focuses on two South Slavic languages, Bulgarian and Macedonian, which are famously exceptional among Slavic languages for having no case declension on non-pronominal nouns or adjectives and for having post-nominal articles. In this paper, I explore a different (and far less recognized) property of these languages that also makes them distinctive, not only among Slavic languages but among most, if not all, natural languages. I trace the historical development of the Bulgarian and Macedonian degree morphemes po- (comparative) and naj- (superlative), which are not only of interest from a traditional philological perspective but furthermore present complex issues with respect to a synchronic analysis of degree adjective morphology. Specifically, I address Bobaljik's (2012) Containment Hypothesis, which states that superlative degree modifiers are morphologically derived directly from comparative degree modifiers. I investigate the Bulgarian and Macedonian degree quantifier series 'many-more-most' – Bg: mnogo-poveče-najmnogo and Mac: mnogu-poveќe-najmnogu – which appear straightforwardly to contradict his claim and provide evidence that this exception arose due to a diachronic shift which may still be underway. [ABSTRACT FROM AUTHOR]

Natural language and visualization are two complementary modalities of human communication that play a crucial role in conveying information effectively. While visualizations help people discover trends, patterns and anomalies in data, natural language descriptions help explain these insights. Thus, combining text with visualizations is a prevalent technique for effectively delivering the core message of the data. Given the rise of natural language generation (NLG), there is a growing interest in automatically creating natural language descriptions for visualizations, which can be used as chart captions, answering questions about charts or telling data‐driven stories. In this survey, we systematically review the state of the art on NLG for visualizations and introduce a taxonomy of the problem. The NLG tasks fall within the domain of natural language interfaces (NLIs) for visualization, an area that has garnered significant attention from both the research community and industry. To narrow down the scope of the survey, we primarily concentrate on the research works that focus on text generation for visualizations. To characterize the NLG problem and the design space of proposed solutions, we pose five Wh‐questions, why and how NLG tasks are performed for visualizations, what the task inputs and outputs are, as well as where and when the generated texts are integrated with visualizations. We categorize the solutions used in the surveyed papers based on these ‘five Wh‐questions’. Finally, we discuss the key challenges and potential avenues for future research in this domain. [ABSTRACT FROM AUTHOR]

Automated summarization of legal texts poses a significant challenge due to the complex and specialized nature of legal documentation. Despite the recent progress in reinforcement learning for natural language text summarization, its application in the legal domain has been less effective. This paper introduces SAC-VAE, a novel reinforcement learning framework specifically designed for legal text summarization. We leverage a Variational Autoencoder (VAE) to condense the high-dimensional state space into a more manageable lower-dimensional feature space. These compressed features are subsequently utilized by the Soft Actor-Critic (SAC) algorithm for policy learning, facilitating the automated generation of summaries from legal texts. Through comprehensive experimentation, we have empirically demonstrated the effectiveness and superior performance of the SAC-VAE framework in legal text summarization. [ABSTRACT FROM AUTHOR]

AbstractObjectiveDesignStudy sampleResultsConclusionEmploying automated language analysis, specifically Meaning Extraction Method (MEM) and Principal Component Analysis (PCA), to identify key factors in open-text responses about hearing aid experiences.Exploratory, cross-sectional design, using an online questionnaire. Responses to a single open-ended question were analysed using MEM, PCA, regression, and correlation analyses.Participants (n = 538) included adult hearing aid users sampled from the Hearing Tracker website community and Lexie Hearing user databases in the United States.The MEM-derived items revealed six factors related to hearing aid experiences: (1) life change, (2) social situation, (3) quality of life, (4) impact and speech understanding, (5) communication and interaction, and (6) music and environmental sounds. IOI-HA item 3 had the most statistically significant correlations with PCA factors. Quantile regression revealed that factors one and two significantly predicted the IOI-HA total score. Positive correlations were observed between self-reported hearing difficulty and factors one, four, and five, as well as between factor one and general health and factor two and physical activity.Natural language analysis of open-ended textual responses can offer valuable insights into hearing aid users’ experiences. Future studies should aim to refine this methodology to enhance clinical relevance and generalisability. [ABSTRACT FROM AUTHOR]

Owing to rapid advancements in information and communication technology, natural language processing (NLP) methods have attracted considerable attention. Estimating uncertainty is a well‐established problem in NLP. Developing systems capable of automatically detecting uncertain terms within a natural language is critical for developing effective text‐analysis algorithms. Consequently, several NLP tools have been developed for this purpose. However, there are several obstacles to the design of fast and effective NLP technologies that effectively analyze ambiguous terms in natural languages. This work focuses on a particular method for detecting uncertainty, termed progressive detection, which can be accomplished using a dynamic probabilistic graphical model to solve these challenges. We iteratively upgraded the Kalman Filter to Labeled Variable Dimension Kalman Filter (LVDKF) based on observations from real datasets. The LVDKF learns two basic features from positive and negative words. The formula was then derived using a dynamic detection system, designed based on conditional probability. Finally, we simulated progressive detection experimental conditions and assessed the LVDKF on two publicly available datasets. It performs better than the baseline methods in these tests, which shows that it is effective at detecting changes over time. [ABSTRACT FROM AUTHOR]

Most of the modern natural language processing (NLP) techniques are based on the vector space models of language, in which each word is represented by a vector in a high dimensional space. One of the earliest successes was demonstrated by the four-term analogical reasoning task: what is to C as B is to A? The trained word vectors form "parallelograms" representing the quadruple of words in analogy. This discovery in NLP offers us insight into our understanding of human semantic representation of words via analogical reasoning. Despite successful applications of the large-scale language models, it has not been fully understood why such parallelograms emerge by learning through natural language data. As the vector space model is not optimized to form parallelograms, the key structure related to geometric shapes of word vectors is expected to be in the data, rather than the models. In the present article, we test our hypothesis that such parallelogram arrangement of word vectors readily exists in the co-occurrence statistics of language. Our approach focuses more on the data itself, and it is different from the existing theoretical approach trying to find the mechanism of parallelogram formation in the algorithms and/or vector arithmetic operations on word vectors. First, our analysis suggested that analogical reasoning is possible by decomposition of the bigram co-occurrence matrix. Second, we demonstrated the formation of a parallelepiped, a more structured geometric object than a parallelogram, by creating a small artificial corpus and its word vectors. With these results, we propose a refined form of distributional hypothesis pointing out an isomorphism between a sort of symmetry or exchangeability and word co-occurrence statistics. [ABSTRACT FROM AUTHOR]

Background: Ontologies are fundamental components of informatics infrastructure in domains such as biomedical, environmental, and food sciences, representing consensus knowledge in an accurate and computable form. However, their construction and maintenance demand substantial resources and necessitate substantial collaboration between domain experts, curators, and ontology experts. We present Dynamic Retrieval Augmented Generation of Ontologies using AI (DRAGON-AI), an ontology generation method employing Large Language Models (LLMs) and Retrieval Augmented Generation (RAG). DRAGON-AI can generate textual and logical ontology components, drawing from existing knowledge in multiple ontologies and unstructured text sources. Results: We assessed performance of DRAGON-AI on de novo term construction across ten diverse ontologies, making use of extensive manual evaluation of results. Our method has high precision for relationship generation, but has slightly lower precision than from logic-based reasoning. Our method is also able to generate definitions deemed acceptable by expert evaluators, but these scored worse than human-authored definitions. Notably, evaluators with the highest level of confidence in a domain were better able to discern flaws in AI-generated definitions. We also demonstrated the ability of DRAGON-AI to incorporate natural language instructions in the form of GitHub issues. Conclusions: These findings suggest DRAGON-AI's potential to substantially aid the manual ontology construction process. However, our results also underscore the importance of having expert curators and ontology editors drive the ontology generation process. [ABSTRACT FROM AUTHOR]

AbstractNatural-language spatial relations between geographic entities (geoentities) reflect diverse perceptions influenced by factors like location, culture, and linguistic conventions. These relations play a crucial role in supporting geospatial tasks, such as question answering and cognitive reasoning. While prior studies focused on a limited set of human-selected spatial terms and geometric attributes, they often overlooked essential semantic attributes. To overcome this limitation, we developed a Spatial Relation-based Knowledge Graph Embedding framework, SR-KGE, with new KG fusion functions to predict spatial relation terms among distinct geoentities. This method not only considers graph structures and the diversity of natural language expressions in the embedding and learning process, but also incorporates geoentity types as a constraint to capture spatial and semantic relations more accurately. Our experiments on two knowledge graph datasets, one small-scale and one large-scale, have both shown its superior performance in spatial relation inference compared to popular KGE models, including TransE, RotatE, and HAKE. We hope our research will advance the classic study of natural language described spatial relations in a more automated and intelligent way. [ABSTRACT FROM AUTHOR]

The common assumption is that genetic information is built on a four-symbol alphabet, i.e., DNA nucleotides, the smallest meaningful blocks of genomes are codon triplets, and the record of genetic information does not contain any asserted symbols playing the role of the space. It is, however, well known that some nucleotides in some codons are redundant. Our study, therefore, tests the alternative scenario. As the same nucleotide may play various semiotic roles, the genomic alphabet actually contains seven semiotic symbols. Consequently, the meaningful fragments of genomes (words) can be of different sizes, and there are asserted symbols in the record of genomic information. If this is true, then, similarly to natural languages, the frequency-range of these genomic words should follow the power-law distribution. The presented hypothesis was tested, in comparison to competitive (codon-based and n-tuple) forms of tokenization, on a wide range of genomic texts. [ABSTRACT FROM AUTHOR]

Health assessments have long been a significant research topic within the field of health psychology. By analyzing the results of subject scales, these assessments effectively evaluate physical and mental health status. Traditional methods, based on statistical analysis, are limited in accuracy due to their reliance on linear scoring methods. Meanwhile, machine learning approaches, despite their potential, have not been widely adopted due to their poor interpretability and dependence on large amounts of training data. Recently, large language models (LLMs) have gained widespread attention for their powerful natural language understanding capabilities, offering a viable solution to these issues. This study investigates the application of LLMs in enhancing physical and mental health assessments, introducing ScaleLLM. ScaleLLM employs language and knowledge alignment to turn LLMs into expert evaluators for health psychology scales. Experimental results indicate that ScaleLLM can improve the accuracy and interpretability of health assessments. [ABSTRACT FROM AUTHOR]

Introduction: In human-robot interaction (HRI), understanding human intent is crucial for robots to perform tasks that align with user preferences. Traditional methods that aim to modify robot trajectories based on language corrections often require extensive training to generalize across diverse objects, initial trajectories, and scenarios. This work presents ExTraCT, a modular framework designed to modify robot trajectories (and behaviour) using natural language input. Methods: Unlike traditional end-to-end learning approaches, ExTraCT separates language understanding from trajectory modification, allowing robots to adapt language corrections to new tasks-including those with complex motions like scooping-as well as various initial trajectories and object configurations without additional end-to-end training. ExTraCT leverages Large Language Models (LLMs) to semantically match language corrections to predefined trajectory modification functions, allowing the robot to make necessary adjustments to its path. This modular approach overcomes the limitations of pre-trained datasets and offers versatility across various applications. Results: Comprehensive user studies conducted in simulation and with a physical robot arm demonstrated that ExTraCT s trajectory corrections are more accurate and preferred by users in 80% of cases compared to the baseline. Discussion: ExTraCT offers a more explainable approach to understanding language corrections, which could facilitate learning human preferences. We also demonstrated the adaptability and effectiveness of ExTraCT in a complex scenarios like assistive feeding, presenting it as a versatile solution across various HRI applications. [ABSTRACT FROM AUTHOR]

Trigger-action platforms (TAPs) enable users without programming experience to personalize the behavior of Internet of Things applications and services through IF-THEN rules. Unfortunately, the arbitrary connection of smart devices and online services, even with simple rules, such as "IF the entrance Netatmo Wheather Station detects a temperature above 30 \({}^{\circ}C\) (\(86^{\circ}F\)) THEN open the shutters in the living room," might expose users to potential security and privacy risks (e.g., the execution of the previous rule might provide an easy entry point for thieves, especially during the summer vacation period). The goal of our research is to make the users capable of understanding and mitigating the threats and risks associated with the execution of IF-THEN rules. To this end, we define a new challenging task, namely generating post hoc justifications of privacy and security risks associated with automation rules, and propose a novel natural language generation strategy based on hybrid prompt learning producing justifications in the form of real-life threat scenarios. The proposed strategy allows for prompt customization with task-specific information, providing contextual details enabling to grasp the nuances and subtleties of the domain language, resulting in more coherent justifications. The experiments conducted on the if-this-then-that (IFTTT) platform show that our method produces effective justifications, improving the explainability of discrete and hybrid prompting methods up to 27% in BLEURT score. The code of the software is publicly available on GitHub1. [ABSTRACT FROM AUTHOR]

The neural boom that has sparked natural language processing (NLP) research throughout the last decade has similarly led to significant innovations in data-to-text (D2T) generation. This survey offers a consolidated view into the neural D2T paradigm with a structured examination of the approaches, benchmark datasets, and evaluation protocols. This survey draws boundaries separating D2T from the rest of the natural language generation (NLG) landscape, encompassing an up-to-date synthesis of the literature, and highlighting the stages of technological adoption from within and outside the greater NLG umbrella. With this holistic view, we highlight promising avenues for D2T research that focus not only on the design of linguistically capable systems but also on systems that exhibit fairness and accountability. [ABSTRACT FROM AUTHOR]

This research focuses on classifying the Hulu and Kuala Banjarese dialects in the prose text “Datu Kandangan and Datu Kartamina”. These dialects represent linguistic variations resulting from geographical, social, and cultural differences among language communities, particularly in South Kalimantan, Indonesia. Language analysis methods such as Python Natural Language Toolkit (NLTK), NumPy, and Latent Dirichlet Allocation (LDA) Visualization (LyLDAvis) were employed to classify the dialects, involving data preprocessing steps like tokenization, punctuation removal, stop word normalization, and stemming. The research findings reveal the superiority of the "Naive Bayes" method over the "Boolean Query," achieving high accuracy in identifying positive examples and classifying texts into Upper and Lower Banjar dialects. The "Naive Bayes" method outperforms the "Boolean Query" with precision and recall values of 0.955563 and 0.956098, while the "Boolean Query" only reaches 0.021416 and 0.146341. This study makes a significant scholarly contribution to understanding language and cultural diversity in South Kalimantan, opening opportunities for further exploration in developing Natural Language Processing (NLP) technology for Indonesian regional languages. [ABSTRACT FROM AUTHOR]

Foundation model has brought unprecedented opportunities and challenges to the military field due to its powerful capabilities in natural language understanding, logical reasoning, and text generation. This paper reviews related technologies of large language models, including architectural and modeling, adaption and fine-tuning, and prompting and learning, as well as the current status of the development of typical multimodal foundation models. It conducts a deep analysis of the characteristics of mainstream domestic and international military application platforms that utilize foundation models, and designs typical task scenarios and capability requirements for military applications of large models. Then, challenges of military foundation models are discussed. Finally, the key common technologies in military applications are summarized. Hopefully, our work can be beneficial for the design of significant military applications based on foundation model technologies and researches of actual combat. [ABSTRACT FROM AUTHOR]

Image captioning aims to automatically generate a natural language description of a given image, and most state‐of‐the‐art models have adopted an encoder–decoder transformer framework. Such transformer structures, however, show two main limitations in the task of image captioning. Firstly, the traditional transformer obtains high‐level fusion features to decode while ignoring other‐level features, resulting in losses of image content. Secondly, the transformer is weak in modelling the natural order characteristics of language. To address theseissues, the authors propose a HIerarchical and Sequential Transformer (HIST) structure, which forces each layer of the encoder and decoder to focus on features of different granularities, and strengthen the sequentially semantic information. Specifically, to capture the details of different levels of features in the image, the authors combine the visual features of multiple regions and divide them into multiple levels differently. In addition, to enhance the sequential information, the sequential enhancement module in each decoder layer block extracts different levels of features for sequentially semantic extraction and expression. Extensive experiments on the public datasets MS‐COCO and Flickr30k have demonstrated the effectiveness of our proposed method, and show that the authors' method outperforms most of previous state of the arts. [ABSTRACT FROM AUTHOR]

Humans excel at forming mental maps of their surroundings, equipping them to understand object relationships and navigate based on language queries. Our previous work SI Maps (Nanwani L, Agarwal A, Jain K, et al. Instance-level semantic maps for vision language navigation. In: 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN). IEEE; 2023 Aug.) showed that having instance-level information and the semantic understanding of an environment helps significantly improve performance for language-guided tasks. We extend this instance-level approach to 3D while increasing the pipeline's robustness and improving quantitative and qualitative results. Our method leverages foundational models for object recognition, image segmentation, and feature extraction. We propose a representation that results in a 3D point cloud map with instance-level embeddings, which bring in the semantic understanding that natural language commands can query. Quantitatively, the work improves upon the success rate of language-guided tasks. At the same time, we qualitatively observe the ability to identify instances more clearly and leverage the foundational models and language and image-aligned embeddings to identify objects that, otherwise, a closed-set approach wouldn't be able to identify. Project Page - [ABSTRACT FROM AUTHOR]

POI data play an important role in various location-based services, including navigation, positioning, and local search applications. However, as cities rapidly develop, a substantial amount of new POI data are generated daily, often accompanied by issues with the quality of their labels. Therefore, there is an urgent need to implement intelligent inference and enhancement processing for POI data labels. Conventional neural network models primarily target balanced data distribution, but they fail to address the issue of imbalanced distribution of POI data labels in terms of quantity. Furthermore, most neural network classification models implicitly learn the semantic knowledge of different categories from training datasets, neglecting the explicit semantic information offered by natural language labels. Considering the above problems, several negative samples are introduced for each input to a positive class, thereby transforming the multi-classification task into a binary classification problem. Simultaneously, POI data labels are introduced to provide explicit semantic information, and the semantic relationship between POI data labels and their names is determined using cross-coding. Experiments demonstrate that the macro − F1 score for the test dataset, which contains 75 different categories of POI data, reaches 0.84. This result surpasses the performance of traditional methods, highlighting the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Large Language Models (LLMs) such as GPT, BART, and Gemini stand at the forefront of Generative Artificial Intelligence, showcasing remarkable prowess in natural language comprehension and task execution. This paper proposes a novel framework developed on the foundation of Llama 2, aiming to bridge the gap between natural language queries and executable code for geospatial analyses within the PyQGIS environment. It empowers non-expert users to leverage GIS technology without requiring deep knowledge of geospatial programming or tools. Through cutting-edge Natural Language Processing (NLP) techniques, including tailored entity recognition and ontology mapping, the framework accurately interprets user intents and translates them into specific GIS operations. Integration of geospatial ontologies enriches semantic comprehension, ensuring precise alignment between user descriptions, geospatial datasets, and geospatial analysis tasks. A code generation module empowered by Llama 2 converts these interpretations into PyQGIS scripts, enabling the execution of geospatial analysis and results visualization. Rigorous testing across a spectrum of geospatial analysis tasks, with incremental complexity, evaluates the framework and the performance of such a system, with LLM at its core. The proposed system demonstrates proficiency in handling various geometries, spatial relationships, and attribute queries, enabling accurate and efficient analysis of spatial datasets. Moreover, it offers robust error-handling mechanisms and supports tasks related to map styling, visualization, and data manipulation. However, it has some limitations, such as occasional struggles with ambiguous attribute names and aliases, which leads to potential inaccuracies in the filtering and retrieval of features. Despite these limitations, the system presents a promising solution for applications integrating LLMs into GIS and offers a flexible and user-friendly approach to geospatial analysis. [ABSTRACT FROM AUTHOR]

This article identifies some situations of linguistic ambiguities and how general logic tries to solve them and it analyzes the influence it can have in particular situations. Since the topic is vast, we stopped at the ambiguities of language caused by identity. First, the attempt to solve the problem was the enunciation of the laws of logic, namely the principle of identity, and we followed its limits. There are analyzed three aspects emphasizing the degree of remaining ambiguity: the relationship between intension and extension, vague terms, and symbolical-metaphorical thinking. Each of the examples given accentuated the limits of classical logic in the face of the natural language ambiguity problems. Ambiguous situations do not only have effects on logical or communicative levels; we also showed the impact on the fields of psycho-social interventions: therapy, counseling, and education. [ABSTRACT FROM AUTHOR]

Videos or Vlogs primarily contain three types of information: natural language, facial movements, and audio activity. These three types of information can effectively portray the publisher's possible attitude. Most recent studies on the multimodal sentiment analysis challenge have focused on two angles. The first is to examine each mode in isolation, but doing so will ignore the interaction between them. The alternative is to integrate information from several modes and treat it all equally, but doing so will add a lot of noise input and disregard critical modes' dominant positions (such as text). The main fusion module based on the mask attention mechanism and the dynamic aggregation attention mechanism, together with the auxiliary fusion module based on the directed cross-modal transformer, can solve these challenges. At the same time, the issue of data misalignment caused by the various sampling rates of each modal sequence is also taken into account in this paper. Our model can deal with the ''unaligned" data, pay greater attention to the crucial modes and features of sentiment information, and reduce the noise input by using the main and auxiliary fusion mechanisms. The model has undergone thorough testing in two publicly accessible multimodal datasets (CMU-MOSI and CMU-MOSEI), and the findings show that our model performs significantly better than state-of-the-art models. [ABSTRACT FROM AUTHOR]

This study focuses on the neural machine translation task for the TR-EN language pair, which is considered a low-resource language pair. We investigated fine-tuning strategies for pre-trained language models. Specifically, we explored the effectiveness of parameter-efficient adapter methods for fine-tuning multilingual pre-trained language models. Various combinations of LoRA and bottleneck adapters were experimented with. The combination of LoRA and bottleneck adapters demonstrated superior performance compared to other methods. This combination required only 5% of the pre-trained language model to be fine-tuned. The proposed method enhances parameter efficiency and reduces computational costs. Compared to the full fine-tuning of the multilingual pre-trained language model, it showed only a 3% difference in the BLEU score. Thus, nearly the same performance was achieved at a significantly lower cost. Additionally, models using only bottleneck adapters performed worse despite having a higher parameter count. Although adding LoRA to pre-trained language models alone did not yield sufficient performance, the proposed method improved machine translation. The results obtained are promising, particularly for low-resource language pairs. The proposed method requires less memory and computational load while maintaining translation quality. [ABSTRACT FROM AUTHOR]

Large Language Models (LLMs) excel in fields such as natural language understanding, generation, complex reasoning, and biomedicine. With advancements in materials science, traditional manual annotation methods for phase diagrams have become inadequate due to their time-consuming nature and limitations in updating thermodynamic databases. To overcome these challenges, we propose a framework based on instruction tuning, utilizing LLMs for automated end-to-end annotation of phase diagrams. High-quality phase diagram images and expert descriptions are collected from handbooks and then preprocessed to correct errors, remove redundancies, and enhance information. These preprocessed data form a golden dataset, from which a subset are used to train LLMs through hierarchical sampling. The fine-tuned LLM is then tested for automated phase diagram annotation. Results show that the fine-tuned model achieves a cosine similarity of 0.8737, improving phase diagram comprehension accuracy by 7% compared to untuned LLMs. To the best of our knowledge, this is the first paper to propose using LLMs for the automated annotation of phase diagrams, replacing traditional manual annotation methods and significantly enhancing efficiency and accuracy. [ABSTRACT FROM AUTHOR]

The phenomenon of bias existed widely in human society, and typically manifested through natural language. Traditional bias studies have mainly focused on static word embedding models, but with the continuous evolution of natural language processing technology, research has gradually shifted towards pre-trained models with stronger contextual processing capabilities. As a further development of pre-trained models, although large language models have been widely deployed in multiple applications due to their remarkable performance and broad prospects, they may still capture social biases from unprocessed training data and propagate these biases to downstream tasks. Biased large language model systems can cause adverse social impacts and other potential harm. Therefore, there is an urgent need for further exploration of bias in large language models. This paper discussed the origins of bias in natural language processing and provided an analysis and summary of the development of bias evaluation and mitigation methods from word embedding models to the current large language models, aiming to provide valuable references for future related research. [ABSTRACT FROM AUTHOR]

Language technologies are gaining momentum as textual information saturates social networks and media outlets, compounded by the growing role of fake news and disinformation. In this context, approaches to represent and analyse public speeches, news releases, social media posts and other types of discourses are becoming crucial. Although there is a large body of literature on text-based machine learning, it tends to focus on lexical and syntactical issues rather than semantic or pragmatic. Being useful, these advances cannot tackle the nuanced and highly context-dependent problems of discourse evaluation that society demands. In this paper, we present IAT/ML, a metamodel and modelling approach to represent and analyse discourses. IAT/ML focuses on semantic and pragmatic issues, thus tackling a little researched area in language technologies. It does so by combining three different modelling approaches: ontological, which focuses on what the discourse is about; argumentation, which deals with how the text justifies what it says; and agency, which provides insights into the speakers' beliefs, desires and intentions. Together, these three modelling approaches make IAT/ML a comprehensive solution to represent and analyse complex discourses towards their understanding, evaluation and fact checking. [ABSTRACT FROM AUTHOR]

Text-based person search aims to align person images with natural language descriptions, which can be widely used in video surveillance field, such as missing person searching and suspect tracking. In this task, extracting distinct representations and aligning them among identities based on descriptions is a crucial yet challenging problem. Most previous methods rely on additional language parsers or vision techniques to identify and select the relevant regions and words from inputs. However, these methods suffer from heavy computation costs and error accumulation. Meanwhile, simply using horizontal segmentation images to obtain local-level features would harm the reliability of models. To address these problems, we first present a novel Simple and Robust Correlation Filtering (SRCF) method which is capable of effectively extracting key clues and aligning discriminative features. Different from previous works, we design two different types of filtering modules (including denoising filters and dictionary filters) to extract essential features and establish multi-modal mappings. Furthermore, despite the SRCF being pretty well, it is still struggling with semantic ambiguity and uni-modal updating. Therefore, we further propose Multi-modal Adaptive Correlation Filtering (MACF) method that adaptively learns the vital regions and keywords with a shared update strategy. Meanwhile, we introduce a new mutually conditional gate to dynamically control the updating process of filters. Extensive experiments demonstrate that both proposed methods improve the robustness and reliability of the model and achieve better performance on the two text-based person search datasets. [ABSTRACT FROM AUTHOR]

Metaphor has long been a topic of investigation in scholarly design research. Studies across various design contexts have demonstrated the value of deliberately exploring and using metaphors to facilitate different phases of the design process. However, the analytical potential of metaphors has often been overshadowed by their generative use, leaving their significance in the research phase of the design process relatively unexplored. This article aims to broaden our understanding of the role metaphors can play in design, delving into the analytical role of metaphors in human-centered design research, particularly focusing on the insights that users' and other stakeholders' metaphors can offer. We present the rationale for adopting metaphor analysis as a method for design research, drawing on conceptual metaphor theory and evidence from analogous research studies. Acknowledging the natural creativity in language use of ordinary people, we argue for the potential this inherent creativity holds for user research and co-creativity settings. We discuss the value of attending to the metaphors in design research, introduce relevant concepts from linguistic metaphor analysis, and demonstrate their practical application on designers' use of user metaphors. We address the advantages and challenges of adopting metaphor analysis as a method for design research and provide recommendations for designers seeking to instrumentalize metaphors in their design processes. [ABSTRACT FROM AUTHOR]

The foundations of AI design discourse are worth analyzing. Here, attention is paid to the nature of theory languages used in designing new AI technologies because the limits of these languages can clarify some fundamental questions in the development of AI. We discuss three types of theory language used in designing AI products: formal, computational, and natural. Formal languages, such as mathematics, logic, and programming languages, have fixed meanings and no actual-world semantics. They are context- and practically content-free. Computational languages use terms referring to the actual world, i.e., to entities, events, and thoughts. Thus, computational languages have actual-world references and semantics. They are thus no longer context- or content-free. However, computational languages always have fixed meanings and, for this reason, limited domains of reference. Finally, unlike formal and computational languages, natural languages are creative, dynamic, and productive. Consequently, they can refer to an unlimited number of objects and their attributes in an unlimited number of domains. The differences between the three theory languages enable us to reflect on the traditional problems of strong and weak AI. [ABSTRACT FROM AUTHOR]

Could translation be fully automated? We must first acknowledge the complexity, ambiguity, and diversity of natural languages. These aspects of natural languages, when combined with a particular dilemma known as the computational dilemma, appear to imply that the machine translator faces certain obstacles that a human translator has already managed to overcome. At the same time, science has not yet solved the problem of how human brains process natural languages and how human beings come to acquire natural language understanding. We will then distinguish between the task of translation and the responsibility of the translator. Thereafter, we will conduct a survey of the methods of machine translation (viz. RBMT, SMT, NMT, foundation models or large language models). These methods will then be critically evaluated both in general and relative to Bar-Hillel's hypothesis about the impossibility of fully automatic, high-quality machine translation (Fahqmt). Some concluding remarks will be made about the scope, prospects, and limits of machine translation. [ABSTRACT FROM AUTHOR]

Biases in the retrieval of personal, autobiographical memories are a core feature of multiple mental health disorders, and are associated with poor clinical prognosis. However, current assessments of memory bias are either reliant on human scoring, restricting their administration in clinical settings, or when computerized, are only able to identify one memory type. Here, we developed a natural language model able to classify text-based memories as one of five different autobiographical memory types (specific, categoric, extended, semantic associate, omission), allowing easy assessment of a wider range of memory biases, including reduced memory specificity and impaired memory flexibility. Our model was trained on 17,632 text-based, human-scored memories obtained from individuals with and without experience of memory bias and mental health challenges, which was then tested on a dataset of 5880 memories. We used 20-fold cross-validation setup, and the model was fine-tuned over BERT. Relative to benchmarking and an existing support vector model, our model achieved high accuracy (95.7%) and precision (91.0%). We provide an open-source version of the model which is able to be used without further coding, by those with no coding experience, to facilitate the assessment of autobiographical memory bias in clinical settings, and aid implementation of memory-based interventions within treatment services. [ABSTRACT FROM AUTHOR]

Understanding spoken language is crucial for conversational agents, with intent detection and slot filling being the primary tasks in natural language understanding (NLU). Enhancing the NLU tasks can lead to an accurate and efficient virtual assistant thereby reducing the need for human intervention and expanding their applicability in other domains. Traditionally, these tasks have been addressed individually, but recent studies have highlighted their interconnection, suggesting better results when solved together. Recent advances in natural language processing have shown that pretrained word embeddings can enhance text representation and improve the generalization capabilities of models. However, the challenge of poor generalization in joint learning models for intent detection and slot filling remains due to limited annotated datasets. Additionally, traditional models face difficulties in capturing both the semantic and syntactic nuances of language, which are vital for accurate intent detection and slot filling. This study proposes a hybridized text representation method using a multichannel convolutional neural network with three embedding channels: non-contextual embeddings for semantic information, part-of-speech (POS) tag embeddings for syntactic features, and contextual embeddings for deeper contextual understanding. Specifically, we utilized word2vec for non-contextual embeddings, one-hot vectors for POS tags, and bidirectional encoder representations from transformers (BERT) for contextual embeddings. These embeddings are processed through a convolutional layer and a shared bidirectional long short-term memory (BiLSTM) network, followed by two softmax functions for intent detection and slot filling. Experiments on the air travel information system (ATIS) and SNIPS datasets demonstrated that our model significantly outperformed the baseline models, achieving an intent accuracy of 97.90% and slot filling F1-score of 98.86% on the ATIS dataset, and an intent accuracy of 98.88% and slot filling F1-score of 97.07% on the SNIPS dataset. These results highlight the effectiveness of our proposed approach in advancing dialogue systems, and paving the way for more accurate and efficient natural language understanding in real-world applications. [ABSTRACT FROM AUTHOR]

Pre‐trained multilingual language models (PMLMs) such as mBERT and XLM‐R have shown good cross‐lingual transferability. However, they are not specifically trained to capture cross‐lingual signals concerning sentiment words. This poses a disadvantage for low‐resource languages (LRLs) that are under‐represented in these models. To better fine‐tune these models for sentiment classification in LRLs, a novel intermediate task fine‐tuning (ITFT) technique based on a sentiment lexicon of a high‐resource language (HRL) is introduced. The authors experiment with LRLs Sinhala, Tamil and Bengali for a 3‐class sentiment classification task and show that this method outperforms vanilla fine‐tuning of the PMLM. It also outperforms or is on‐par with basic ITFT that relies on an HRL sentiment classification dataset. [ABSTRACT FROM AUTHOR]

Large language models have achieved outstanding performance on various downstream tasks with their advanced understanding of natural language and zero-shot capability. However, they struggle with knowledge constraints, particularly in tasks requiring complex reasoning or extended logical sequences. These limitations can affect their performance in question answering by leading to inaccuracies and hallucinations. This paper proposes a novel framework called KnowledgeNavigator that leverages large language models on knowledge graphs to achieve accurate and interpretable multi-hop reasoning. Especially with an analysis-retrieval-reasoning process, KnowledgeNavigator searches the optimal path iteratively to retrieve external knowledge and guide the reasoning to reliable answers. KnowledgeNavigator treats knowledge graphs and large language models as flexible components that can be switched between different tasks without additional costs. Experiments on three benchmarks demonstrate that KnowledgeNavigator significantly improves the performance of large language models in question answering and outperforms all large language models-based baselines. [ABSTRACT FROM AUTHOR]

Natural language inference (NLI) is a fundamental and quite challenging task in natural language processing, requiring efficient methods that are able to determine whether given hypotheses derive from given premises. In this paper, we apply explainability techniques to natural-language-inference methods as a means to illustrate the decision-making procedure of its methods. First, we investigate the performance and generalization capabilities of several transformer-based models, including BERT, ALBERT, RoBERTa, and DeBERTa, across widely used datasets like SNLI, GLUE Benchmark, and ANLI. Then, we employ stacking-ensemble techniques to leverage the strengths of multiple models and improve inference performance. Experimental results demonstrate significant improvements of the ensemble models in inference tasks, highlighting the effectiveness of stacking. Specifically, our best-performing ensemble models surpassed the best-performing individual transformer by 5.31% in accuracy on MNLI-m and MNLI-mm tasks. After that, we implement LIME and SHAP explainability techniques to shed light on the decision-making of the transformer models, indicating how specific words and contextual information are utilized in the transformer inferences procedures. The results indicate that the model properly leverages contextual information and individual words to make decisions but, in some cases, find difficulties in inference scenarios with metaphorical connections which require deeper inferential reasoning. [ABSTRACT FROM AUTHOR]