<?xml version="1.0" encoding="utf-8"?>
    <rss version="2.0">
      <channel xmlns:content="http://purl.org/rss/1.0/modules/content/">
        <title>Frontiers in Artificial Intelligence | Language and Computation section | New and Recent Articles</title>
        <link>https://www.frontiersin.org/journals/artificial-intelligence/sections/language-and-computation</link>
        <description>RSS Feed for Language and Computation section in the Frontiers in Artificial Intelligence journal | New and Recent Articles</description>
        <language>en-us</language>
        <generator>Frontiers Feed Generator,version:1</generator>
        <pubDate>2026-05-02T12:05:56.894+00:00</pubDate>
        <ttl>60</ttl>
        <item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1708566</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1708566</link>
        <title><![CDATA[Hybrid artificial intelligence architectures for automatic text correction in the Kazakh language]]></title>
        <pubdate>2025-12-12T00:00:00Z</pubdate>
        <category>Brief Research Report</category>
        <author>Laura Baitenova</author><author>Saule Tussupova</author><author>Saken Mambetov</author><author>Gauhar Munaitbas</author><author>Gulnar Mukhamejanova</author>
        <description><![CDATA[The Kazakh language, as an agglutinative and morphologically rich language, presents significant challenges for the development of natural language processing (NLP) tools. Traditional rule-based analyzers provide full coverage but lack flexibility; statistical and neural models handle disambiguation more effectively, yet require large annotated corpora and substantial computational resources. This paper presents a hybrid morphological analyzer that integrates Finite-State Transducers (FST), Conditional Random Fields (CRF), and transformer-based architectures (KazRoBERTa, mBERT). For the experiments, a new corpus, KazMorphCorpus-2025, was created, consisting of 150,000 sentences from diverse domains annotated for morphological analysis. Experimental evaluation demonstrated that the KazRoBERTa model consistently outperforms mBERT in terms of accuracy, F1-score, and prediction speed. The hybrid architecture effectively combines the exhaustive coverage of FST with the contextual disambiguation of neural networks, reducing errors associated with homonymy, borrowings, and long affixal chains. The results confirm that the proposed system achieves a balance between accuracy, efficiency, and scalability. The study underscores the practical significance of hybrid approaches for tasks such as spell checking, information retrieval, and machine translation in the Kazakh language, as well as their potential transferability to other low-resource Turkic languages. Future work will include the expansion of the corpus, integration of KazBERT and mBERT models, and validation of the proposed approach in applied NLP systems.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1648073</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1648073</link>
        <title><![CDATA[Crowdsourcing lexical diversity]]></title>
        <pubdate>2025-12-05T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Hadi Khalilia</author><author>Jahna Otterbacher</author><author>Gábor Bella</author><author>Shandy Darma</author><author>Fausto Giunchiglia</author>
        <description><![CDATA[Lexical-semantic resources (LSRs), such as online lexicons and wordnets, are fundamental to natural language processing applications as well as to fields such as linguistic anthropology and language preservation. In many languages, however, such resources suffer from quality issues: incorrect entries, incompleteness, but also the rarely addressed issue of bias toward the English language and Anglo-Saxon culture. Such bias manifests itself in the absence of concepts specific to the language or culture at hand, the presence of foreign (Anglo-Saxon) concepts, as well as in the lack of an explicit indication of untranslatability, also known as cross-lingual lexical gaps, when a term has no equivalent in another language. This paper proposes a novel crowdsourcing methodology for reducing bias in LSRs. Crowd workers compare lexemes from two languages, focusing on domains rich in lexical diversity, such as kinship or food. Our LingoGap crowdsourcing platform facilitates comparisons through microtasks identifying equivalent terms, language-specific terms, and lexical gaps across languages. We validated our method by applying it to two case studies focused on food-related terminology: (1) English and Arabic, and (2) Standard Indonesian and Banjarese. These experiments identified 2,140 lexical gaps in the first case study and 951 in the second. The success of these experiments confirmed the usability of our method and tool for future large-scale lexicon enrichment tasks.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1666074</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1666074</link>
        <title><![CDATA[A geometric semantic model and Parts-of-Sense Inference annotation framework]]></title>
        <pubdate>2025-11-13T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Kiran Pala</author><author>S. Shalu</author><author>Vasudevan Nedumpozhimana</author><author>Kamal Kumar Choudhary</author>
        <description><![CDATA[We introduce a geometric semantic model designed to capture fine-grained semantic representations in a multidimensional space. Building on this model, we develop a novel annotation framework that facilitates detailed semantic analysis across languages. Central to our approach is a set of Parts-of-Sense Inference (POSI) tags: 135 interpretable four-letter codes that annotate subtle semantic attributes often overlooked by traditional models. To evaluate the cross-linguistic and cross-structural applicability of this framework, we annotate expressions in four typologically diverse languages. Our results demonstrate that the proposed model provides an interpretable, cognitively plausible approach to semantic representation and can serve as a robust tool for investigating language processing and meaning inference across linguistic contexts.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1678043</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1678043</link>
        <title><![CDATA[Detection of cloned voices in realistic forensic voice comparison scenarios]]></title>
        <pubdate>2025-11-13T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Pedro Univaso</author><author>Eugenia San Segundo</author>
        <description><![CDATA[Deepfakes and synthetic audio significantly degrade the performance of automatic speaker recognition systems commonly used in forensic laboratories. We investigate the effectiveness of Mel-Frequency Cepstral Coefficients (MFCCs) for detecting cloned voices, ultimately concluding that MFCC-based methods are insufficient as a universal anti-spoofing tool due to their inability to generalize across different cloning algorithms. Furthermore, we evaluate the performance of the HIVE AI-deepfake Content Detection tool, noting its vulnerability to babble noise and signal saturation, which are common in real-world forensic recordings. This investigation emphasizes the ongoing competition between voice cloning and detection technologies, underscoring the urgent need for more robust and generalized anti-spoofing systems for forensic applications.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1623573</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1623573</link>
        <title><![CDATA[Weaponizing cognitive bias in autonomous systems: a framework for black-box inference attacks]]></title>
        <pubdate>2025-08-20T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Shiyong Chu</author><author>Yuwei Chen</author>
        <description><![CDATA[Autonomous systems operating in high-dimensional environments increasingly rely on prioritization heuristics to allocate attention and assess risk, yet these mechanisms can introduce cognitive biases such as salience, spatial framing, and temporal familiarity that influence decision-making without altering the input or accessing internal states. This study presents Priority Inversion via Operational Reasoning (PRIOR), a black-box, non-perturbative diagnostic framework that employs structurally biased but semantically neutral scenario cues to probe inference-level vulnerabilities without modifying pixel-level, statistical, or surface semantic properties. Given the limited accessibility of embodied vision-based systems, we evaluate PRIOR using large language models (LLMs) as abstract reasoning proxies to simulate cognitive prioritization in constrained textual surveillance scenarios inspired by Unmanned Aerial Vehicle (UAV) operations. Controlled experiments demonstrate that minimal structural cues can consistently induce priority inversions across multiple models, and joint analysis of model justifications and confidence estimates reveals systematic distortions in inferred threat relevance even when inputs are symmetrical. These findings expose the fragility of inference-level reasoning in black-box systems and motivate the development of evaluation strategies that extend beyond output correctness to interrogate internal prioritization logic, with implications for dynamic, embodied, and visually grounded agents in real-world deployments.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1609097</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1609097</link>
        <title><![CDATA[An overview of model uncertainty and variability in LLM-based sentiment analysis: challenges, mitigation strategies, and the role of explainability]]></title>
        <pubdate>2025-08-11T00:00:00Z</pubdate>
        <category>Review</category>
        <author>David Herrera-Poyatos</author><author>Carlos Peláez-González</author><author>Cristina Zuheros</author><author>Andrés Herrera-Poyatos</author><author>Virilo Tejedor</author><author>Francisco Herrera</author><author>Rosana Montes</author>
        <description><![CDATA[Large Language Models (LLMs) have significantly advanced sentiment analysis, yet their inherent uncertainty and variability pose critical challenges to achieving reliable and consistent outcomes. This paper systematically explores the Model Variability Problem (MVP) in LLM-based sentiment analysis, characterized by inconsistent sentiment classification, polarization, and uncertainty arising from stochastic inference mechanisms, prompt sensitivity, and biases in training data. We present illustrative examples and two case studies to highlight its impact and analyze the core causes of MVP, discussing a dozen fundamental reasons for model variability. We pay especial atenttion to explainabily, with an analysis of its importance in LLMs from the MVP perspective. In addition, we investigate key challenges and mitigation strategies, paying particular attention to the role of temperature as a driver of output randomness and highlighting the crucial role of explainability in improving transparency and user trust. By providing a structured perspective on stability, reproducibility, and trustworthiness, this study helps develop more reliable, explainable, and robust sentiment analysis models, facilitating their deployment in high-risk domains such as finance, healthcare and policy making, among others.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1592013</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1592013</link>
        <title><![CDATA[Large language models for closed-library multi-document query, test generation, and evaluation]]></title>
        <pubdate>2025-08-06T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Claire Randolph</author><author>Adam Michaleas</author><author>Darrell O. Ricke</author>
        <description><![CDATA[IntroductionLearning complex, detailed, and evolving knowledge is a challenge in multiple technical professions. Relevant source knowledge is contained within many large documents and information sources with frequent updates to these documents. Knowledge tests need to be generated on new material and existing tests revised, tracking knowledge base updates. Large Language Models (LLMs) provide a framework for artificial intelligence-assisted knowledge acquisition and continued learning. Retrieval-Augmented Generation (RAG) provides a framework to leverage available, trained LLMs combined with technical area-specific knowledge bases.MethodsHerein, two methods are introduced (DaaDy: document as a dictionary and SQAD: structured question answer dictionary), which together enable effective implementation of LLM-RAG question-answering on large documents. Additionally, the AI for knowledge intensive tasks (AIKIT) solution is presented for working with numerous documents for training and continuing education. AIKIT is provided as a containerized open source solution that deploys on standalone, high performance, and cloud systems. AIKIT includes LLM, RAG, vector stores, relational database, and a Ruby on Rails web interface.ResultsCoverage of source documents by LLM-RAG generated questions decreases as the length of documents increase. Segmenting source documents improve coverage of generated questions. The AIKIT solution enabled easy use of multiple LLM models with multimodal RAG source documents; AIKIT retains LLM-RAG responses for queries against one or multiple LLM models.DiscussionAIKIT provides an easy-to-use set of tools to enable users to work with complex information using LLM-RAG capabilities. AIKIT enables easy use of multiple LLM models with retention of LLM-RAG responses.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1619489</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1619489</link>
        <title><![CDATA[A multidimensional comparison of ChatGPT, Google Translate, and DeepL in Chinese tourism texts translation: fidelity, fluency, cultural sensitivity, and persuasiveness]]></title>
        <pubdate>2025-07-24T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Shiyue Chen</author><author>Yan Lin</author>
        <description><![CDATA[This study systematically compares the translation performance of ChatGPT, Google Translate, and DeepL on Chinese tourism texts, focusing on two prompt-engineering strategies. Using a mixed-methods approach that combines quantitative expert assessments with qualitative analysis, the evaluation centers on fidelity, fluency, cultural sensitivity, and persuasiveness. ChatGPT outperformed its counterparts across all metrics, especially when culturally tailored prompts were used. However, it occasionally introduced semantic shifts, highlighting a trade-off between accuracy and rhetorical adaptation. Despite its strong performance, human post-editing remains necessary to ensure semantic precision and professional standards. The study demonstrates ChatGPT’s potential in domain-specific translation tasks while calling for continued oversight in culturally nuanced content.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1650320</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1650320</link>
        <title><![CDATA[Correction: Stylistic variation across English translations of Chinese science fiction: Ken Liu versus ChatGPT]]></title>
        <pubdate>2025-07-03T00:00:00Z</pubdate>
        <category>Correction</category>
        <author>Frontiers Production Office </author>
        <description></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1576750</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1576750</link>
        <title><![CDATA[Stylistic variation across English translations of Chinese science fiction: Ken Liu versus ChatGPT]]></title>
        <pubdate>2025-06-19T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Pingdi Zhou</author><author>Jiajun Cheng</author>
        <description><![CDATA[Advancements in computational tools, including neural machine translation (NMT) and large language models (LLMs), have revolutionized literary stylistics and opened new avenues in corpus-based translation studies (CBTS). Yet, the style of LLM-produced translations, especially in science fiction (SF) literature, remain understudied. This study examines stylistic variation across English translations of Chinese SF by translator Ken Liu and ChatGPT-4o. Thirteen works translated by both were compared using Multi-Dimensional analysis on key dimensions. Stylometric tests assessed within-translator and between-translator variations, and functional analysis interpreted the subordinate linguistic features. Findings reveal that Ken Liu adapts his style to each story’s depth, exhibiting greater variation, while GPT maintains a more consistent style. Ken Liu’s less narrative style enhances resonance through a minimalist approach, whereas GPT’s more narrative style offers clarity but may undermine thematic impact. The study contributes to CBTS by providing a methodological framework for comparing human and LLM translations in terms of style. It highlights a collaborative model that combines human creativity with LLM efficiency, necessitating continuous upskilling among students, educators, and practitioners to adapt to LLMs’ growing presence in translation. Ultimately, by exploring the intersection of linguistics, literature, and artificial intelligence, the study pushes the boundaries of translation studies and practices.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1582287</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1582287</link>
        <title><![CDATA[Who speaks next? Multi-party AI discussion leveraging the systematics of turn-taking in Murder Mystery games]]></title>
        <pubdate>2025-06-17T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Ryota Nonomura</author><author>Hiroki Mori</author>
        <description><![CDATA[IntroductionMulti-agent systems utilizing large language models (LLMs) have shown great promise in achieving natural dialogue. However, smooth dialogue control and autonomous decision making among agents still remain challenging.MethodsIn this study, we focus on conversational norms such as adjacency pairs and turn-taking found in conversation analysis and propose a new framework called “Murder Mystery Agents” that applies these norms to AI agents' dialogue control. As an evaluation target, we employed the “Murder Mystery” game, a reasoning-type table-top role-playing game that requires complex social reasoning and information manipulation. The proposed framework integrates next speaker selection based on adjacency pairs and a self-selection mechanism that takes agents' internal states into account to achieve more natural and strategic dialogue.ResultsTo verify the effectiveness of this new approach, we analyzed utterances that led to dialogue breakdowns and conducted automatic evaluation using LLMs, as well as human evaluation using evaluation criteria developed for the Murder Mystery game. Experimental results showed that the implementation of the next speaker selection mechanism significantly reduced dialogue breakdowns and improved the ability of agents to share information and perform logical reasoning.DiscussionThe results of this study demonstrate that the systematics of turn-taking in human conversation are also effective in controlling dialogue among AI agents, and provide design guidelines for more advanced multi-agent dialogue systems.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1581129</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1581129</link>
        <title><![CDATA[On automatic decipherment of lost ancient scripts relying on combinatorial optimisation and coupled simulated annealing]]></title>
        <pubdate>2025-05-30T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Fabio Tamburini</author>
        <description><![CDATA[This paper introduces a novel method for addressing the challenge of deciphering ancient scripts. The approach relies on combinatorial optimisation along with coupled simulated annealing, an advanced technique for non-convex optimisation. Encoding solutions through k-permutations facilitates the representation of null, one-to-many, and many-to-one mappings between signs. In comparison to current state-of-the-art systems evaluated on established benchmarks from literature and three new benchmarks introduced in this study, the proposed system demonstrates superior performance in enhancing cognate identification results.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1456245</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1456245</link>
        <title><![CDATA[Exploring the evolution and future prospects of Amharic to English machine translation: a systematic review]]></title>
        <pubdate>2025-05-23T00:00:00Z</pubdate>
        <category>Systematic Review</category>
        <author>Muluken Hussen Asebel</author><author>Shimelis Getu Assefa</author><author>Mesfin Abebe Haile</author>
        <description><![CDATA[IntroductionIn the last couple of decades, Amharic-English translation has greatly improved from a rule-based approach to contemporary systems that apply neural networks. Even after these advancements, problems remain because of the Amharic language’s resource-scarce nature, such as inadequate datasets, tools for working with the language, and the intricate semantics and grammar of Amharic as compared to English. This systematic review seeks to analyze the evolution of the Amharic-English machine translation, the prominent ongoing difficulties, the noteworthy research undertakings, and the prospects of the research focus.MethodsThis review uses a systematic approach to study the literature on Amharic-English machine translation. Important documents were retrieved from academic websites, and those with relevance to the methodologies of machine translation, language resources development, and evaluation practices were chosen. Primarily, the focus was on both statistical and neural machine translation models, especially those with transformer structures.ResultsThe initial attempts to translate English to Amharic and vice-versa relied on statistic machine translation (SMT), which set the stage for the evolution to neural machine translation (NMT). The use of transformer models has impacted the accuracy and fluidity of translations tremendously. Still, there is a lack of sufficient parallel corpora, effective methods for tokenization of Amharic, and other resources. Recently, the focus has been on creating new datasets, improving token-level engineering, and modifying NMT models for Amharic’s complex morphological structure.DiscussionThe complete solutions for enhancing Amharic-English translation remain elusive and include the lack of sufficient data, semantic correspondence, and grammatical consistency within and across translations. Pursuable avenues include augmentation of data, tokenization on the language level, and incorporation of linguistic elements into the parallel corpora. In addition, creating effective evaluation frameworks along with comprehensive linguistic data is important for assessing and improving translation tools. With these changes, cross-cultural interaction and increasing accessibility to modern technologies will be achieved.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1509338</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1509338</link>
        <title><![CDATA[Making sense of transformer success]]></title>
        <pubdate>2025-04-01T00:00:00Z</pubdate>
        <category>Hypothesis and Theory</category>
        <author>Nicola Angius</author><author>Pietro Perconti</author><author>Alessio Plebe</author><author>Alessandro Acciai</author>
        <description><![CDATA[This article provides an epistemological analysis of current attempts of explaining how the relatively simple algorithmic components of neural language models (NLMs) provide them with genuine linguistic competence. After introducing the Transformer architecture, at the basis of most of current NLMs, the paper firstly emphasizes how the central question in the philosophy of AI has been shifted from “can machines think?”, as originally put by Alan Turing, to “how can machines think?”, pointing to an explanatory gap for NLMs. Subsequently, existing explanatory strategies for the functioning of NLMs are analyzed to argue that they, however debated, do not differ from the explanatory strategies used in cognitive science to explain intelligent behaviors of humans. In particular, available experimental studies turned to test the theory of mind, discourse entity tracking, and property induction in NLMs are examined under the light of the functional analysis in the philosophy of cognitive science; the so-called copying algorithm and the induction head phenomenon of a Transformer are shown to provide a mechanist explanation of in-context learning; finally, current pioneering attempts to use NLMs to predict brain activation patterns when processing language are here shown to involve what we call a co-simulation, in which a NLM and the brain are used to simulate and understand each other.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1558696</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1558696</link>
        <title><![CDATA[Gender and content bias in Large Language Models: a case study on Google Gemini 2.0 Flash Experimental]]></title>
        <pubdate>2025-03-18T00:00:00Z</pubdate>
        <category>Brief Research Report</category>
        <author>Roberto Balestri</author>
        <description><![CDATA[This study evaluates the biases in Gemini 2.0 Flash Experimental, a state-of-the-art large language model (LLM) developed by Google, focusing on content moderation and gender disparities. By comparing its performance to ChatGPT-4o, examined in a previous work of the author, the analysis highlights some differences in ethical moderation practices. Gemini 2.0 demonstrates reduced gender bias, notably with female-specific prompts achieving a substantial rise in acceptance rates compared to results obtained by ChatGPT-4o. It adopts a more permissive stance toward sexual content and maintains relatively high acceptance rates for violent prompts (including gender-specific cases). Despite these changes, whether they constitute an improvement is debatable. While gender bias has been reduced, this reduction comes at the cost of permitting more violent content toward both males and females, potentially normalizing violence rather than mitigating harm. Male-specific prompts still generally receive higher acceptance rates than female-specific ones. These findings underscore the complexities of aligning AI systems with ethical standards, highlighting progress in reducing certain biases while raising concerns about the broader implications of the model's permissiveness. Ongoing refinements are essential to achieve moderation practices that ensure transparency, fairness, and inclusivity without amplifying harmful content.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1523336</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1523336</link>
        <title><![CDATA[Determining the meter of classical Arabic poetry using deep learning: a performance analysis]]></title>
        <pubdate>2025-02-14T00:00:00Z</pubdate>
        <category>Methods</category>
        <author>A. M. Mutawa</author><author>Ayshah Alrumaih</author>
        <description><![CDATA[The metrical structure of classical Arabic poetry, deeply rooted in its rich literary heritage, is governed by 16 distinct meters, making its analysis both a linguistic and computational challenge. In this study, a deep learning-based approach was developed to accurately determine the meter of Arabic poetry using TensorFlow and a large dataset. Character-level encoding was employed to convert text into integers, enabling the classification of both full-verse and half-verse data. In particular, the data were evaluated without removing diacritics, preserving critical linguistic features. A train–test–split method with a 70–15–15 division was utilized, with 15% of the total dataset reserved as unseen test data for evaluation across all models. Multiple deep learning architectures, including long short-term memory (LSTM), gated recurrent units (GRU), and bidirectional long short-term memory (Bi-LSTM), were tested. Among these, the bidirectional long short-term memory model achieved the highest accuracy, with 97.53% for full-verse and 95.23% for half-verse data. This study introduces an effective framework for Arabic meter classification, contributing significantly to the application of artificial intelligence in natural language processing and text analytics.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2024.1490698</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2024.1490698</link>
        <title><![CDATA[Language writ large: LLMs, ChatGPT, meaning, and understanding]]></title>
        <pubdate>2025-02-12T00:00:00Z</pubdate>
        <category>Hypothesis and Theory</category>
        <author>Stevan Harnad</author>
        <description><![CDATA[Apart from what (little) OpenAI may be concealing from us, we all know (roughly) how Large Language Models (LLMs) such as ChatGPT work (their vast text databases, statistics, vector representations, and huge number of parameters, next-word training, etc.). However, none of us can say (hand on heart) that we are not surprised by what ChatGPT has proved to be able to do with these resources. This has even driven some of us to conclude that ChatGPT actually understands. It is not true that it understands. But it is also not true that we understand how it can do what it can do. I will suggest some hunches about benign “biases”—convergent constraints that emerge at the LLM scale that may be helping ChatGPT do so much better than we would have expected. These biases are inherent in the nature of language itself, at the LLM scale, and they are closely linked to what it is that ChatGPT lacks, which is direct sensorimotor grounding to connect its words to their referents and its propositions to their meanings. These convergent biases are related to (1) the parasitism of indirect verbal grounding on direct sensorimotor grounding, (2) the circularity of verbal definition, (3) the “mirroring” of language production and comprehension, (4) iconicity in propositions at LLM scale, (5) computational counterparts of human “categorical perception” in category learning by neural nets, and perhaps also (6) a conjecture by Chomsky about the laws of thought. The exposition will be in the form of a dialogue with ChatGPT-4.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2025.1477246</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2025.1477246</link>
        <title><![CDATA[Analysis of argument structure constructions in the large language model BERT]]></title>
        <pubdate>2025-01-31T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Pegah Ramezani</author><author>Achim Schilling</author><author>Patrick Krauss</author>
        <description><![CDATA[Understanding how language and linguistic constructions are processed in the brain is a fundamental question in cognitive computational neuroscience. In this study, we investigate the processing and representation of Argument Structure Constructions (ASCs) in the BERT language model, extending previous analyses conducted with Long Short-Term Memory (LSTM) networks. We utilized a custom GPT-4 generated dataset comprising 2000 sentences, evenly distributed among four ASC types: transitive, ditransitive, caused-motion, and resultative constructions. BERT was assessed using the various token embeddings across its 12 layers. Our analyses involved visualizing the embeddings with Multidimensional Scaling (MDS) and t-Distributed Stochastic Neighbor Embedding (t-SNE), and calculating the Generalized Discrimination Value (GDV) to quantify the degree of clustering. We also trained feedforward classifiers (probes) to predict construction categories from these embeddings. Results reveal that CLS token embeddings cluster best according to ASC types in layers 2, 3, and 4, with diminished clustering in intermediate layers and a slight increase in the final layers. Token embeddings for DET and SUBJ showed consistent intermediate-level clustering across layers, while VERB embeddings demonstrated a systematic increase in clustering from layer 1 to 12. OBJ embeddings exhibited minimal clustering initially, which increased substantially, peaking in layer 10. Probe accuracies indicated that initial embeddings contained no specific construction information, as seen in low clustering and chance-level accuracies in layer 1. From layer 2 onward, probe accuracies surpassed 90 percent, highlighting latent construction category information not evident from GDV clustering alone. Additionally, Fisher Discriminant Ratio (FDR) analysis of attention weights revealed that OBJ tokens had the highest FDR scores, indicating they play a crucial role in differentiating ASCs, followed by VERB and DET tokens. SUBJ, CLS, and SEP tokens did not show significant FDR scores. Our study underscores the complex, layered processing of linguistic constructions in BERT, revealing both similarities and differences compared to recurrent models like LSTMs. Future research will compare these computational findings with neuroimaging data during continuous speech perception to better understand the neural correlates of ASC processing. This research demonstrates the potential of both recurrent and transformer-based neural language models to mirror linguistic processing in the human brain, offering valuable insights into the computational and neural mechanisms underlying language understanding.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2024.1236310</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2024.1236310</link>
        <title><![CDATA[A framework for extending co-creative communication models to sustainability research]]></title>
        <pubdate>2024-08-05T00:00:00Z</pubdate>
        <category>Hypothesis and Theory</category>
        <author>Guanhong Li</author><author>Xiaoyun Guo</author>
        <description><![CDATA[The UN Sustainable Development Goals (SDGs) present a challenge due to their potential for conflicting objectives, which hinders their effective implementation. In order to address the complexity of sustainability issues, a framework capable of capturing the specificity of diverse sustainability issues while offering a common methodology applicable across contexts is required. Co-creative communication can be regarded as a key source of uncertainty within functional systems, as it can be instrumental in realizing and sustaining sustainability. In this regard, the studies in Constructive approaches to Co-creative Communication (CCC), particularly those employing artificial intelligence (AI) methodologies such as computational social science and innovation studies, hold significant value for both theoretical and applied sustainability research. However, existing CCC frameworks cannot be directly applied to sustainability research. This work bridges this gap by proposing a framework that outlines a general approach to establishing formalized definitions of sustainability from the lens of communication. This approach enables the direct application of CCC models to sustainability studies. The framework is based on systems theory and the methodologies of artificial intelligence, including computational/symbolic modeling and formal methods. This framework emphasizes the social function of co-creative communication and the interaction between the innovation process and the sustainability of the system. It can be concluded that the application of our framework enables the achievements of CCC to be directly applied to sustainability research. Researchers from different disciplines are therefore able to establish their own specific definitions of sustainability, which are tailored to their particular concerns. Our framework lays the groundwork for future sustainability studies that employs CCC, facilitating the integration of CCC insights into sustainability research and application. The outcomes of computational creativity research based on AI technologies, such as distributed artificial intelligence and self-organizing networks, can deepen the understanding of sustainability mechanisms and drive their practical applications. Furthermore, the functional role of co-creative communication in societal sustainability proposed in this work offers a novel perspective for future discussions on the evolutionary adaptation of co-creative communication.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2024.1287877</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2024.1287877</link>
        <title><![CDATA[Exploring the performance of automatic speaker recognition using twin speech and deep learning-based artificial neural networks]]></title>
        <pubdate>2024-02-08T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Julio Cesar Cavalcanti</author><author>Ronaldo Rodrigues da Silva</author><author>Anders Eriksson</author><author>Plinio A. Barbosa</author>
        <description><![CDATA[This study assessed the influence of speaker similarity and sample length on the performance of an automatic speaker recognition (ASR) system utilizing the SpeechBrain toolkit. The dataset comprised recordings from 20 male identical twin speakers engaged in spontaneous dialogues and interviews. Performance evaluations involved comparing identical twins, all speakers in the dataset (including twin pairs), and all speakers excluding twin pairs. Speech samples, ranging from 5 to 30 s, underwent assessment based on equal error rates (EER) and Log cost-likelihood ratios (Cllr). Results highlight the substantial challenge posed by identical twins to the ASR system, leading to a decrease in overall speaker recognition accuracy. Furthermore, analyses based on longer speech samples outperformed those using shorter samples. As sample size increased, standard deviation values for both intra and inter-speaker similarity scores decreased, indicating reduced variability in estimating speaker similarity/dissimilarity levels in longer speech stretches compared to shorter ones. The study also uncovered varying degrees of likeness among identical twins, with certain pairs presenting a greater challenge for ASR systems. These outcomes align with prior research and are discussed within the context of relevant literature.]]></description>
      </item>
      </channel>
    </rss>