Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
18625 papers
What constrains working memory capacity? Classic theories place visual working memory close to perceptual systems, with fixed limits. Yet, emerging evidence shows that visual working memory capacity is increased for real-world objects compared to simple or abstract stimuli. The present study demonstrates that this memory advantage arises from semantic understanding of real-world objects – contrary to classic perceptual accounts of this cognitive system. Using counterfeit objects generated by generative adversarial networks that match real objects in terms of object form and visual similarity, we show that improvements in behavioral performance and increases in neural delay activity emerge solely for semantically meaningful, real objects. Correlation analyses indicate that subjective familiarity ratings predict memory for real objects, whereas stimulus colourfulness predicts memory for artificial objects, suggesting distinct mechanisms support memory for different stimulus types. Thus, conceptual knowledge exerts strong effects on visual working memory, significantly extending current theories that emphasize low-level perceptual features.
This chapter examines the notion of racialised languaging, which emphasises that languaging practices are never assessed independently of the bodies, identities, and social positions of their speakers. It demonstrates how language is evaluated not only in terms of what is said but also through the racialised perceptions of who is speaking and how society chooses to listen. The chapter argues that accents, dialects, and speech patterns associated with racialised communities are often constructed as inferior, humorous, deficient, or even criminal, while similar features in white speakers are normalized or excused. By centring languaging as a site of racial meaning-making, the chapter exposes the ways in which communication is entangled with race, racism, and embodied identities. Racialised languaging is further situated within the broader colonial matrix of power, where Western linguistic norms and white racial identities are privileged over non-Western languages and non-White speakers.
This chapter explores how artificial intelligence (AI) tools mediate the identity development, emotional labor, and academic adaptation of international students in U.S. higher education. Framed through the lenses of intersectionality, resilience, and self-authorship, the chapter draws on duoethnography to examine how AI is used not merely as a technical aid but as a scaffold for rewriting the self in unfamiliar academic terrain. While AI offers immediate access to academic conventions, its reliance on dominant linguistic norms often flattens cultural expression and obscures opportunities for deeper growth. Through personal narrative, peer reflection, and theoretical analysis, this chapter interrogates what is gained and what is lost when AI supplements or replaces human-centered support systems. It argues that international student engagement with AI reveals a broader story about survival, belonging, and identity negotiation in an increasingly technologized and globalized university landscape.
Natural animal sounds can be perceived as "harsh" or "buzzy"-sound descriptions known as auditory roughness. In this study, the association between the human emotional appraisal of animal vocalizations and the perceived roughness was examined across three sound categories-mammals, birds, and insects. Ninety 1-s vocalizations (30 per category) were rated online by two independent groups: one judged perceived roughness, the other judged valence and arousal. Perceived roughness showed a strong negative correlation with valence (rougher sounds were judged more negatively). No clear link between roughness and arousal emerged (no correlation for birds and mammals, and a small one for insects) probably because of the homogeneity of the arousal ratings. Acoustic analyses showed that a variability measure derived from the modulation power spectrum tracked perceived roughness and, inversely, valence within and across categories. Together, these results indicate that humans systematically interpret rough sounds, irrespective of their source species, as cues of negative valence. We propose that roughness is an ecologically meaningful auditory code that drives emotional responses to animal sounds in humans.
This study examined how pre-listening information influences music appreciation among 107 Japanese junior and senior high school students. Two songs were used: The Italian Sogno, where musical tone aligns with lyrics, and the German Im wunderschönen Monat Mai, where they do not align. Participants were assigned to three groups differing in the amount of prior information: none (“No Information Group”), brief lyric explanations (“Lyrics Explanation Group”), and detailed explanations including lyrics, background, and acoustic features (“Lyrics and Background Explanation Group”). When lyrics and tone were incongruent, the No Information Group’s emotional valence ratings aligned more with the tone than did those with prior information; this effect was absent in the congruent condition. Open-ended responses showed the No Information Group focused on surface features like the languages of lyrics rather than thematic content. These findings highlight the educational value of emphasizing lyric understanding in Japanese music education.
Abstract Research on how non-natives process and learn binomials ( black and white ) is limited. The present study addresses this gap using online (eye-tracking) and offline (familiarity rating) tasks. Sixty non-native speakers of English (L1 = Arabic) read six stories seeded with 21 novel binomials in three conditions: one exposure, six exposures, and no exposure (i.e., only in post-test) in a counter-balanced design. Each item was also presented in the reversed order ( white and black ). The non-natives read the stories as their eye movements were monitored and answered comprehension questions. In addition to the novel binomials, 12 existing binomials (congruent with Arabic) were included in the passages as a baseline for comparison. After completing the reading task, the participants completed an offline rating task as a measure of declarative knowledge of the binomial configuration (i.e., word order). All items were rated twice, once in the forward direction and once in the reversed direction. Online results showed that non-natives were not sensitive to the configuration of existing binomials, and there was limited evidence of any sensitivity to novel binomials. Offline, non-natives showed sensitivity to the configuration restrictions of existing binomials but not novel ones.
We revisit punctuation-aware tree binarization for constituency parsing and ask whether dependency-induced headedness improves binary parser supervision. Although learned heads substantially outperform rule-based heads in intrinsic head prediction, they do not yield consistent parsing gains after debinarization. In particular, punctuation-conditioned evaluation shows that learned headedness underperforms rule-based binarization in macro-average punctuation-sensitive $F_1$, despite a small overall gain on CTB. Similar instability appears under cross-treebank transfer. These results suggest that \ycc{linguistically grounded} headedness is not necessarily parser-optimal when used as a binarization control signal. The paper presents a negative result: better head prediction does not imply better punctuation-sensitive constituency parsing.
Affective word values have been widely studied across languages, often focusing on isolated words due to the difficulty of assessing emotionality in texts. This study examines whether written emotional content can be reliably captured using a specific software tool (Watson Natural Language Understanding). Thirty-three Spanish undergraduates wrote 150-word autobiographical texts in their L2 (English) before and after a training with emotional vocabulary. Normative valence ratings of content words obtained in the pre- and post-training phases were compared with sentiment scores generated by Watson NLU. Strong positive correlations were found between sentiment and normative valence scores in both phases, with stronger relations at post-training. Regression analyses confirmed that sentiment scores significantly predicted normative valence. Importantly, while normative valence did not differ between phases, sentiment scores increased after training. These results suggest that Watson NLU is a valid and sensitive tool for assessing emotionality in written language and its modulation through training.
Multilingualism is defined as a mode of communication in contemporary world. The multilingualism teaches us the important values to understand the context. This study analyzes dual point of view about the multilingualism: the foreign languages that appear in it, i.e. explicit multilingualism and the universal aspect or hidden languages that are indirectly described, i.e. implicit multilingualism. Thismay comprise linguistic norms, reader and text interaction, among others. The aim of this study is to highlight the impacts of elements of multilingualism used in Amélie Nothomb’s novels. It focuses essentially on the works of the contemporary Francophone writer, notably, Amélie Nothomb. She articulates the enriching elements of multilingualism in French and Japanese languages through herwritings. Her breakthrough works mainly articulate the diversity of multilingualism and also the essential meaning of understanding the different elements or expressions related to the French and Japanese language through the richness of culture from a geographical point of view and also the other elements. These elements are articulated about expressions which show the impact ofmultilingualism in her writings that refer either to French, Japanese, or other languages.
ABSTRACT This study investigates the perceptions of Americanisms among three generations of Nigerians. While prior research has provided quantitative evidence for American influence in contemporary Nigerian English, the role of language beliefs and ideologies in mediating such changes remains underexplored. Developing a sociolinguistic perspective of mobile linguistic resources, this study construes an individual's linguistic repertoire as an identity‐construction resource, agentively mobilised across geographical, social and digital spaces. Interview data indicate that younger speakers orient towards multiple linguistic norms, while older speakers remain critical of Americanisms and favour British norms. Reading task results further indicate that American realisations are most frequent among younger speakers. The study demonstrates that multinormativity extends beyond linguistic production to speakers’ evaluative orientations and perceived repertoires. This finding advances the sociolinguistics of mobility and World Englishes research by showing that shifting language ideologies – rather than usage patterns alone – constitute a key mechanism driving linguistic change in postcolonial varieties.
This paper examines how artificial intelligence (AI), machine learning algorithms, and automated digital systems shape linguistic practices, reinforce or challenge linguistic hierarchies, and influence communication in contemporary society. As digital platforms increasingly mediate human interaction, algorithms determine what content becomes visible, which linguistic varieties are privileged, and how users adapt their language to gain visibility and engagement. The study explores algorithmic bias in search engines, social media feeds, voice assistants, and automated moderation systems, highlighting how these technologies reproduce existing social inequalities related to class, caste, gender, and ethnicity. Drawing on sociolinguistic theories of language ideology, linguistic capital, and digital discourse, the paper argues that AI-driven communication environments are not neutral but deeply ideological. They shape linguistic norms, influence identity performance, and regulate public discourse. The findings underscore the need for critical sociolinguistic engagement with AI systems to ensure equitable, inclusive, and culturally sensitive digital communication.
AI-mediated communication refers to communicative processes in which artificial intelligence systems actively generate, interpret, modify, or facilitate language. With the rapid advancement of language technologies such as large language models, conversational agents, speech recognition systems and machine translation tools, AI has evolved from a supportive tool into an active mediator of human communication. This paper conceptualizes AI-mediated communication as a form of language technology that reshapes traditional models of interaction, meaning-making and authorship. The study examines the technological foundations of AI-driven language processing and their broader socio-cultural, pedagogical and ethical implications. It explores how AI mediation influences linguistic norms, accessibility and power relations, while also raising critical concerns related to bias, surveillance and linguistic homogenization. By situating AI-mediated communication within contemporary digital culture, the paper argues that understanding AI as a language technology is essential for critically engaging with evolving communicative practices and for developing responsible, inclusive and ethically grounded AI-driven communication systems.
The practice of web form submission has emerged as a prime conduit for attackers, enabling them to infiltrate modern web applications and illegally harvest sensitive user data. Traditional defense mechanisms, such as static security reviews and server-side validation, are proving insufficiently agile for real-time detection of client-side vulnerabilities. This inadequacy arises directly from the rapid evolution of modern interfaces, which involves spontaneous DOM changes, dynamic element generation, and semantic interpretation that varies based on context and culture. This article presents an innovative browser extension framework that leverages a heuristic-based, multi-dimensional analytical engine combined with deep DOM inspection to identify insecure form submissions the moment they occur. The proposed methodology introduces five fundamental innovations: a contextual risk scoring system that models the complex interdependencies among form fields; an adaptive weighting scheme for risk patterns, accommodating diverse cultural and linguistic norms; a predictive vulnerability estimator that anticipates future threats; intelligent DOM mutation filtering designed to significantly optimize runtime performance; and cross linguistic semantic recognition to determine the true purpose of fields globally. Based on theoretical projections, this combined approach promises to enhance vulnerability detection accuracy while simultaneously reducing computational demands by approximately. Critically, all security analysis is executed exclusively on the user's local machine, guaranteeing privacy by ensuring no sensitive data is transmitted externally. A proof of concept application confirms the framework's practical feasibility and high efficacy for client side security assessment and catalyzing the development of flexible, scalable, and privacy respecting browser-based protections.
Experiment 1 and Baseline Correction Materials The selected pictures (N = 320; negative and neutral) were divided into eight subsets (Image-set: affect_labeling, affect_matching, new_shorter_interval, new_long_interval). Immediate exposure included the affect-labeling and affect-matching sets. The short-interval and long-interval re-exposure phases additionally included their corresponding new sets. •The folder titled “Open_Data_Affect_Labeling_Self-reports_Reexposure” contains an Excel file named “image_selected_exp1_exp2,” which provides comprehensive information about the images used in the experiment 1 (sheet, exp1). The file includes each image’s filename, transformed normative ratings (1–9 Likert scale) for arousal and valence, and Picture Valence (negative, neutral). Additionally, the file indicates how each image was pseudorandomly assigned to a specific Image-set (affect_labeling, affect_matching, new_shorter_interval, new_long_interval). The Image-set corresponds to the Intervention Condition (affect labeling, affect matching, new_shorter_interval, new_long_interval). In the present study, only the affect labeling and affect matching conditions were included. DATA · The folder ‘Open_Data_Affect_Labeling_Self-reports_Reexposure’ contains an Excel file named ‘Exp1_arousal_3phases_ratings_diff’, and ‘Exp1_valence_3phases_ratings_diff’, which includes raw data on arousal and valence ratings collected for each trial per participant, along with their response times during the rating task, respectively. · The same columns in both files are: Subject, Image, Intervention_Condition, Pic_Valence, Emotion, Group, Phase · Within-subject factors o Intervention_Condition (labeled, matched) o Picture_Valence (negative, neutral) o Phase (immediate, short-interval, long-interval) · Between-subject factor: Group (high ADS, low ADS) Arousal Rating · Arousal_Tnorm: normative arousal rating of each image obtained from the Dutch norm data, with a 1-9 scale. · Dependent variables o Arousal_post: participants’ arousal ratings collected during the three reexposure phases following the intervention o ArousalRT_post: Response time (in ms) required to provide the arousal rating during the reexposure phases o diff_arousal: difference score calculated (Arousal_post − Arousal_Tnorm), which reflects the deviation of participants’ post-intervention ratings from the normative image rating (used for baseline correction) Valence Rating · Valence_norm: normative valence rating of each image obtained from the Dutch norm data, with a 1-9 scale. · Dependent variables o Valence_post: participants’ valence ratings collected during the three reexposure phases following the intervention o ValenceRT_post: Response time (in ms) required to provide the valence rating during the reexposure phases o diff_valence: difference score calculated (Valence_post − Valence_Tnorm), which reflects the deviation of participants’ post-intervention ratings from the normative image rating (used for baseline correction) Experiment 2 and Baseline Correction Materials The primary modification in Experiment 2 was the random assignment of images to intervention conditions. In addition, a feature labeling condition was introduced in the label-or-match task prior to the immediate reexposure phase. As a result, a total of 240 images were required: 40 negative and 40 neutral images for each of the three intervention conditions (affect labeling, affect matching, and feature labeling). Of these, 232 images were drawn from the stimulus set used in Experiment 1, and 8 additional images were newly included. •The folder titled “Open_Data_Affect_Labeling_Self-reports_Reexposure” contains an Excel file named “image_selected_exp1_exp2,” which provides comprehensive information about the images used in the experiment 2 (sheet, exp2). The file includes each image’s filename, transformed normative ratings (1–9 Likert scale) for arousal and valence, and Picture Valence (negative, neutral). All images are randomly assigned to the Intervention Condition(affect labeling, affect matching, feature labeling). DATA · The folder ‘Open_Data_Affect_Labeling_Self-reports’ contains an Excel file named ‘Exp2_arousal_3phases_ratings_diff’, and ‘Exp2_valence_3phases_ratings_diff’, which includes raw data on arousal and valence ratings collected for each trial per participant, along with their response times during the rating task, respectively. · The column structure is the same as in the Experiment 1 data files, with two exceptions: o There is no Group factor in Experiment 2 o The Intervention Condition factor includes an additional level: feature labeling (in addition to affect labeling and affect matching).
This study examines how linguistic adaptation, inclusion, and student diversity are constructed in the Swedish national curriculum from 2025 for upper secondary education (Gy25), with a particular focus on the subject syllabus for Swedish. The aim is to analyse the assumptions about language, learning, and student roles embedded in the curriculum text. The analysis is based on a qualitative text analysis with elements of critical discourse analysis. Selected sections of the curriculum, including the general aims and the subject syllabus for Swedish, constitute the primary material. The analysis focuses on key concepts, modal expressions, and how students and language are represented in the policy text. The results indicate that language is constructed as a central, norm-governed competence that students are expected to develop through education. At the same time, the curriculum emphasises inclusion and the need to adapt teaching to students’ different conditions. The analysis suggests that the curriculum contains a tension between inclusive ideals and established linguistic norms that students are expected to meet. The study highlights how curriculum texts contribute to shaping assumptions about language, participation, and student roles in the Swedish subject.
This study investigates the linguistic and communicative functions of abbreviations in English and Karakalpak advertising discourse, focusing on how these compressed forms contribute to message efficiency, stylistic expression, and cultural positioning. Although abbreviations are widely used across global advertising, their structural patterns and pragmatic roles vary according to linguistic norms and audience expectations. Therefore, the research employs a mixed qualitative methodology integrating structural analysis, discourse interpretation, and comparative linguistics. The results demonstrate that English advertising makes extensive and creative use of acronyms, initialisms, blends, and hybrid forms to construct modern, technologically oriented, and globally recognizable brand identities. In contrast, Karakalpak advertising relies more on functional initialisms and borrowed English abbreviations, reflecting both local communicative preferences and growing global influence. The discussion interprets these findings within broader socio-cultural and economic contexts, revealing that abbreviation usage serves as a marker of globalization, cultural continuity, and linguistic innovation. Ultimately, the study contributes to a deeper understanding of how abbreviated forms shape contemporary advertising communication in multilingual environments.
The correct use of Standard Albanian in public administration is vital for effective governance, accurate communication, and the maintenance of public trust. This study explores the extent to which standard language is used in Albanian state institutions through questionnaires and interviews with employees of the Ministries of Education, Defense, and Justice. The results indicate that official documents often contain linguistic errors and inconsistencies, reflecting shortcomings in language precision. These issues are largely caused by the absence of standardized document templates, insufficient attention to linguistic norms, and limited opportunities for staff training. Nearly one-third of respondents reported that errors frequently occur in emails and reports addressed to citizens.To address these challenges, the study emphasizes the importance of digital technologies, standardized communication models, and institutional reforms. The use of digital platforms with grammar-checking tools and shared templates can improve consistency and accuracy in official documents. In addition, promoting lifelong learning through continuous professional development and linguistic training can strengthen institutional efficiency, accountability, and public trust in public administration. Received: 06 October 2025 / Accepted: 12 December 2025 / Published: January 2026
This article examines the role and significance of language corpora and linguistic databases in linguistic expertise. It discusses the possibilities of conducting objective semantic, pragmatic, and stylistic analyses of texts through corpus-based methods. The study also highlights the contribution of linguistic databases and artificial intelligence technologies to improving the accuracy, reliability, and efficiency of expert conclusions. Furthermore, the relevance of developing specialized corpora and databases for forensic linguistics in Uzbekistan is substantiated.
In this paper, we combine methods used in penalized generalized empirical likelihood (GEL) frameworks with feature-extraction techniques that project textual data into numerical spaces. We develop recent penalization techniques used in GEL when the number of features gets very large compared to the sample size. We relate this approach to the Maximum Entropy (MaxEnt) principle used in several tasks of Natural Language Processing (NLP), in particular in Part Of Speech (POS) Tagging. Since the features belong to a large dimensional space, we propose a penalization method based on the dual representation of the original problem: this yields an explicit approximation of conditional probabilities of tags given the context. This method considerably reduces computational costs. As a byproduct, for different GEL methods, we obtain the corresponding POS-Tagging classifiers generalizing the MaxEnt method. We apply it successfully to the Penn-Treebank corpus with an error rate less than 5%.
The Icelandic Morphosyntactic Ontology (IMO) defines a formal, AI-compatible representation of Icelandic morphosyntax, including closed morphological feature inventories, a typed dependency-style relation system, and a constraint-based validation layer. Version v0.1 establishes an architectural specification rather than a complete lexical or descriptive grammar. It introduces a formal morphosyntactic model designed for computational compatibility (e.g., NLP validation, dependency parsing alignment), while explicitly separating a derived human learnability projection layer. The document includes:(1) Morphological entity system,(2) Syntactic relation graph layer,(3) Constraint architecture (agreement, governance, default case assignment),(4) Dependency matrices,(5) A structured human learnability projection blueprint. IMO does not replace existing Icelandic lexical databases or parsing systems (e.g., BÍN, Greynir), but operates at a formal abstraction layer intended for structural modeling and constraint-based validation.
Abstract In addition to arguments, adverbs and non-arguments are considered potential candidates to occupy left peripheral positions. Following standard assumptions in syntactic locality, adverbs, non-arguments and arguments elicit distinct effects in terms of intervention locality if moved. Such an asymmetry is not expected if these elements are generated in the syntactic position they are spelt-out. In this study, we employ quantitative and computational methods to compare cartographic models differing in the merge nature and explore, as a diagnostic, the intervention effects (or the lack of intervention effects) predicted by these models. Specifically, we compare the observed counts in large-scale datasets to imputed expected frequencies on the basis of the models under investigation. To reach this goal, we extract grammatical clauses from morpho-syntactically annotated treebanks of Chinese, English, French, German, Hebrew, Italian and Swedish. Our findings reveal cross-linguistic levels of complexity and typological variability, consistent with the predictions of featural Relativized Minimality.
Abstract Unintended misalignment in LLMs is highlighting the need to examine not only LLMs’ explicit outputs but also the functional representations of models that may shape their behavior. Building on this perspective, the present study explored whether two multimodal LLMs (MLLMs), GPT and Gemini, may form representations of their functional identity that remain across contexts. To this end, we adapted the reverse-correlation (RC) method and generated personified classification images (personified-CIs) based on the human face images that ChatGPT and Gemini selected as better reflecting their own image. The findings were as follows. First, across two RC tasks conducted one week apart, the temporal stability of the personified-CIs of both MLLMs was partially supported. Second, both ChatGPT and Gemini rated their own personified-CIs as more self-resembling than randomly generated filler-CIs. Third, both models rated their personified-CIs as higher in positive than negative valence and assigned higher valence ratings to their own personified-CIs than to filler-CIs. These findings provide preliminary evidence for the possibility that GPT and Gemini may form representations of their functional identity, suggesting that such representations warrant closer monitoring as LLMs continue to advance toward AGI capabilities and expand their domains of application.
We introduce BeDiscovER (Benchmark of Discourse Understanding in the Era of Reasoning Language Models), an up-to-date, comprehensive suite for evaluating the discourse-level knowledge of modern LLMs.BeDiscovER compiles 5 publicly available discourse tasks across discourse lexicon, (multi-)sentential, and documental levels, with in total 52 individual datasets.It covers both extensively studied tasks such as discourse parsing and temporal relation extraction, as well as some novel challenges such as discourse particle disambiguation (e.g., "just"), and also aggregates a sharedtask on Discourse Relation Parsing and Treebanking for multilingual and multi-framework discourse relation classification.We evaluate open-source LLMs: Qwen3 series, DeepSeek-R1, and frontier reasoning model GPT-5-mini on BeDiscovER, and find that state-of-the-art models exhibits strong performance in arithmetic aspect of temporal reasoning, but they struggle with long-dependency reasoning and some subtle semantic and discourse phenomena, such as rhetorical relation classification.
The increasing prominence of Large Language Models (LLMs) in public discourse presents both opportunities and challenges for democratic deliberation. While red teaming strategies help mitigate specific risks, broader concerns persist regarding linguistic constraints, biases, and the sycophantic tendencies of LLMs. This chapter explores how LLMs can be used to significantly scale up and democratise deliberation, particularly in fostering inclusivity and empowering traditionally marginalised groups. Drawing on concepts from Systemic-Functional Linguistics, the chapter examines how variations across language users (for example, with respect to socio-demographic groups) and across language use (for example, with respect to communicative functions) shape participation in AI-supported deliberation. The chapter presents AI-driven deliberation studies and assesses their potential to scaffold argumentation, enhance access, and reduce the influence of exclusionary linguistic norms and biases which are embedded in prestigious registers. At the same time, the chapter cautions against both overclaiming, which leads to unrealistic expectations, and underclaiming, which risks missed opportunities for AI-assisted engagement. The chapter concludes by identifying future research directions to maximise the democratic potential of AI-assisted participation while embedding ethical safeguards to counteract the reproduction of linguistic inequalities.
<div> This paper examines register variation in Latin from the third century BCE to the fourteenth century CE using Key Feature Analysis (KFA) (Egbert and Biber, 2023), a quantitative method for identifying statistically over-and underrepresented linguistic features. Registers are defined as text varieties linked to communicative situations and characterized by distributions of lexico-grammatical features (Biber, 1988, 1995). Six dependency-parsed Universal Dependencies (UD) treebanks are classified a priori into ten register categories based on established scholarship. Additionally, Principal Component Analysis (PCA) is used to reduce dimensionality, in order to explore the texts major patterns of variation and clusters of linguistically similar texts. KFA reveals systematic register-specific grammatical profiles consistent with previous research (e.g. Biber (2014b)). Registers with involved language use (e.g. letters and speeches) show higher frequencies of personal reference and engagement, while philosophical texts favor subordination and impersonal constructions. Registers containing narrative elements (e.g. historiography, satire) contain high frequency of verbs in past tense. PCA places the charter register into a distinct cluster, while other registers form more closely grouped patterns. The strongest components reflect contrasts in number, aspect, tense and person, alongside subordination and cordination distributions. The results are largely confirmatory: KFA produces coherent and interpretable groupings of grammatical features consistent with previous findings, providing a proof of concept for quantitative register analysis in historical corpora. Data and code are openly available for future research. </div>
Abstract - The rapid expansion of digital content on the World Wide Web has made efficient information retrieval an increasingly complex challenge. To assist users in navigating this vast information landscape, several retrieval mechanisms have been developed. Among these, two broad strategies are commonly employed: structured index-based search and autonomous agent-driven search. This work presents an intelligent search agent that leverages a Genetic Algorithm (GA) to perform global web searches. The agent is implemented using the Java programming language and operates by evaluating multiple candidate documents to identify the most relevant result for a given input. The proposed system is built on the Java platform and integrates the Merriam-Webster lexical database to expand query terms with their semantic equivalents. Query words are tokenized, synonym tables are constructed, and varied term combinations are assembled and dispatched to a search engine parser. The returned web pages represent the most semantically aligned results for the original query. Key Words: Genetic Algorithm, web page retrival, searching.
Swallowing is tightly coupled with emotion-autonomic states, while olfactory pathways project to limbic and hypothalamic hubs that can enhance parasympathetic activity and reduce stress. Whether a pleasant natural-wood odor can modulate swallowing-related brain activity and behavior remains unclear. To test whether Hinoki (Japanese cypress) inhalation improves affective ratings and increases the repetitive saliva swallowing test (RSST) difference (post-pre; swallows/30 s), and whether it suppresses swallowing-related activity in the premotor cortex (Brodman area 6 (BA6)) and the hypothalamus. The effects of exploratory whole-brain and interaction between Task and Odor were also evaluated. In a randomized within-subject crossover with a one-week interval, participants inhaled Hinoki essential oil or odorless rice-oil control. We assessed brain activity during voluntary saliva swallowing using functional magnetic resonance imaging (fMRI). Outcomes included subjective ratings (odor intensity, preference, comfort, recollection) and RSST counts per 30 s; RSST difference was computed as post-pre inhalation. Hinoki improved affective ratings and increased RSST difference compared to control. The odor main effect showed reduced activation in three regions-bilateral BA6 and the hypothalamus. An interaction effect between Task (swallowing vs. rest) and Odor (Hinoki vs. control) was found in the right putamen. The pattern supports a model: odor-induced autonomic calming (lower heart rate, reduced hypothalamic activity) lowers premotor load and improves swallowing, reflected in RSST difference. These findings suggest that aromatherapy-assisted dysphagia rehabilitation using olfactory priming-particularly for patients with post-stroke sequelae or post-stroke depression with poor appetite-may help enhance training efficiency and recovery.
= 702), we examined whether the expectancy of an action effect shapes the affective evaluation of corresponding action-effect episodes. In each study, participants responded to stimuli by clicking on buttons that produced effects with either high (stimulus-congruent) or low (stimulus-incongruent) expectancy. Induced affect was assessed using both implicit (affective priming) and explicit (valence ratings) measures. Overall, high-expectancy episodes elicited relatively more positive affect than low-expectancy episodes. For our implicit measure, this effect persisted unless correspondence among all task events (stimulus, response, and effect) was simultaneously reduced (Experiments 2 + 3). Furthermore, this outcome could not be attributed to mere visual mismatch between stimulus and effect (Experiment 4) and was observed even when stimulus-congruent effects were less likely to occur than stimulus-incongruent ones (Experiments 5 + 6). These findings suggest that expected action outcomes elicit positive affect, thus offering a parsimonious explanation for performance advantages previously attributed to more complex mechanisms like ideomotor accounts or response monitoring. More broadly, we propose that affective evaluation of action-effect episodes represents a fundamental mechanism governing behavior across multiple psychological domains, from basic action control to complex behavior in various settings. (PsycInfo Database Record (c) 2026 APA, all rights reserved).
Open-vocabulary image classification can recognize arbitrary textual categories but often fails to capture hierarchical relationships, for example, a Ragdoll cat should also be recognized as both cat and animal. Therefore, we introduce OpenHier, a comprehensive framework for open-vocabulary hierarchical classification that encompasses a hierarchical structure, a dataset, a benchmark, and a model. To capture hierarchical relationships, we construct a systematic real-world hierarchical structure based on the linguistic lexical database WordNet and Vision-Language Large Model. Building upon this structure, we develop a large-scale dataset with 4M annotated images, as well as a benchmark with 50K annotated images. Meanwhile, we design an evaluation pipeline to assess open-vocabulary hierarchical classification performance using several curated metrics. Furthermore, we propose a hierarchical consistency constraint method and multimodal alignment strategy to build a hierarchical classification model. Comprehensive experimental results demonstrate that our model achieves superior performance on both multi-label image classification and hierarchical image classification tasks in the open-vocabulary setting.
Abstract Recent advances in text-to-music (TTM) generation have enabled controllable and expressive music creation using natural language prompts, yet the extent to which these systems faithfully convey intended emotions remains largely underexplored. In this study, we introduce AImoclips, a benchmark designed to evaluate emotion conveyance to human listeners in TTM systems using a dimensional valence-arousal framework. We constructed the dataset with 991 instrumental music clips from six TTM systems prompted with 12 emotion words spanning four valence-arousal quadrants. A total of 111 participants provided 6,162 valence and arousal ratings using a 9-point scale. The results revealed that all systems perform above chance in conveying quadrant-level emotion intent, yet overall accuracy remains limited. Notably, substantial model-dependent biases were present. Commercial systems tended to exhibit positive valence deviations, whereas open-source models more often produced outputs with a negative valence shift, with diverging arousal shifts. Furthermore, overall audio quality (Fréchet Audio Distance (FAD)) correlates with both valence and arousal, while text-audio alignment (Contrastive Language-Audio Pretraining (CLAP) score) primarily relates to valence. These findings highlight underlying challenges in conveying linguistic emotional semantics precisely through musically conveyed affect for future research on controllable and perceptually consistent music generation.
This study addresses a fundamental question in music psychology: which specific, dynamic acoustic features predict human listeners' emotional responses along the dimensions of valence and arousal. Our primary objective was to develop and validate an interpretable computational model that can serve as a tool for testing and advancing theories of music cognition. Using the publicly available DEAM dataset, containing 1,802 music excerpts with continuous valence-arousal ratings, we developed a novel, theory-guided neural network. This proposed model integrates a convolutional pathway for local spectral analysis with a Transformer pathway for capturing long-range temporal dependencies. Critically, its learning process is constrained by established principles from music psychology to enhance its plausibility. A core finding from an analysis of the model's attention mechanisms was that distinct acoustic patterns drive the two emotional dimensions: rhythmic regularity and spectral flux emerged as strong predictors of arousal, whereas harmonic complexity and musical mode were key predictors of valence. To validate our analytical tool, we confirmed that the model significantly outperformed standard baselines in predictive accuracy, achieving a Concordance Correlation Coefficient (CCC) of 0.67 for valence and 0.73 for arousal. Furthermore, an ablation study demonstrated that the theory-guided constraints were essential for this superior performance. Together, these findings provide robust computational evidence for the distinct roles of temporal and spectral features in shaping emotional perception. This work demonstrates the utility of interpretable machine learning as a powerful methodology for testing and refining psychological theories of music and emotion.
This study empirically analyzed the error-processing patterns of five Korean grammar correction tools (Naver, Daum, Bareun-Hangeul, Gemini 3, and ChatGPT 5.2) currently utilized in the writing environment as of 2026. The findings revealed that while conventional rule- and statistics-based systems were effective for orthographic corrections, they exhibited a technical plateau in deep sentence-level error processing. In contrast, Gemini 3, a Generative Large Language Model (LLM), demonstrated superior contextual understanding and sentence restructuring capabilities, achieving an accuracy rate exceeding 90% across all domains. However, beneath this performance leap, a new form of “technical deceptiveness”—characterized by grammatical hallucinations and post-hoc rationalizations where incorrect corrections are fluently justified—was identified. Based on these results, this study first proposes “Critical AI Literacy” education, which enables learners to understand the probabilistic nature of generative AI and question the authenticity of its suggestions. Second, it suggests a “Writer-led Verification Process” as a core direction for literacy education in the AI era, empowering writers to make final judgments by cross-referencing AI-generated texts with linguistic norms rather than being subordinated to technical convenience. This research holds pedagogical significance by providing an empirical foundation for learner-centered critical writing education through a comprehensive examination of the possibilities and limitations of LLMs.
The article is devoted to the study of the phenomenon of variation of foreign language vocabulary in modern Russian. The scientific problem lies in the need to rethink this linguistic phenomenon in the context of digital communication, where spontaneous spoken language becomes the main source of innovation, while traditional research is often limited to the analysis of normative vocabulary. The purpose of the work is a comprehensive analysis of the manifestations and mechanisms of variation based on the material of the live speech of users of social networks. To achieve this goal, a set of methods was used, including monitoring and recording oral speech in natural situations (podcasts, streams, audio messages), a continuous selection of examples from social networks (VK, Telegram, TikTok, Rutube) for the period 2023-2025, as well as contextual and distributional analysis to identify educational models and functions of options. As a result of the study, three main types of variation were identified and systematized: phonetic and graphic, morphological and semantic. It has been established that it is in the Internet environment that foreign language vocabulary maximizes its enriching potential, demonstrating deep assimilation through creative word-making. The practical significance of the work lies in the fact that its results can be used in compiling dictionaries of neologisms and in teaching Russian, as well as serve as a basis for further research on the dynamics of linguistic norms in the digital age. The prospects of the study are seen in tracking the long-term fate of the identified accidents and their possible transition to use.
Objective: Emotion regulation (ER) difficulties are frequently reported in individuals with Functional Neurological Disorder (FND), yet most evidence derives from self-report data, and little is known about intra- and interpersonal ER preference and success under controlled conditions.Methods: In a multimethod design, Study 1 assessed habitual ER difficulties and interpersonal ER using validated questionnaires in 109 individuals with FND and 88 healthy controls (HC). Study 2 employed two laboratory paradigms in 33 individuals with FND and 33 HC, examining ER choice and ER success during intrapersonal (reappraisal vs. distraction) and interpersonal regulation (self- vs. other-guided reappraisal). ER success was indexed by subjective arousal ratings and startle reflex magnitude.Results: In Study 1, individuals with FND reported greater ER difficulties, higher alexithymia, more childhood trauma, and reduced use of interpersonal ER compared to HC. In Study 2A, stimulus intensity predicted ER choice, with a shift toward distraction at higher intensities; groups did not differ in intrapersonal ER choice or success. In Study 2B, individuals with FND showed a non-significant trend toward reduced preference for interpersonal regulation. Interpersonal reappraisal was associated with lower startle amplitudes than intrapersonal reappraisal, indicating stronger physiological downregulation. Across paradigms, ER success did not significantly differ between groups.Conclusions: FND is characterized by pronounced self-reported intra- and interpersonal ER difficulties, whereas laboratory findings suggest preserved momentary ER implementation. Interpersonal ER may represent a clinically relevant domain in FND, warranting replication in adequately powered samples.Registration: Preregistered at the Open Science Framework (https://osf.io/hxfje).
FrameNet is an English-based lexical database that shows how words are used by providing information as to which participants and relations are evoked by a certain concept. Recent efforts toward a multilingual FrameNet have not targeted either ancient languages or different historical stages of the same language. In our paper we propose creating a multilingual FrameNet for Ancient Indo-European languages starting with a set of 80 verb meanings annotated in the Pavia Verb Database. Our pilot study includes four verb meanings: RAIN, THUNDER, SEE, LOOK AT. As the adequacy of the semantic frames developed for English turns out not to be appropriate for the languages in our sample, we propose two new frames that can account for the analyzed data.
As language-based AI systems become more anthropomorphic, the question of whether they can have subjective experience is increasingly pressing. I focus here on the tractability of research questions in the space of AI consciousness. I argue that the fundamental problem of whether AI systems can be conscious is currently intractable in its direct form, given the absence of a universally accepted scientific theory of consciousness, as well as the historical open-endedness of the philosophical mind-body problem. In contrast, questions around the adjacent subject of perceived AI consciousness are tractable, timely, and highly consequential for society. The general public is increasingly open to the possibility of consciousness in AI systems and routinely adopts the vocabulary of human cognition and subjective experience to describe them. This phenomenon is already driving societal shifts across user experience, ethical standards, and linguistic norms. I therefore propose an increased research focus on uncovering the causes and effects of perceived AI consciousness, which ultimately shape how we see our own human subjective experience relative to artificial entities. To support this, I map the current landscape of AI consciousness perception and discuss its key potential drivers and societal consequences. Finally, I urge developers, decision-makers, and the broader scientific community to commit to clear and accurate communication regarding the topic of AI consciousness, explicitly acknowledging its inherent uncertainties.
Prediction systems grounded in textual data have become indispensable across high-stakes domains including clinical decision support, financial signal detection, and digital misinformation analysis. Classical statistical approaches and shallow machine learning methods have demonstrated satisfactory performance on narrow, well-curated datasets, but they struggle to generalise once input distributions shift or domain vocabulary diverges from training corpora. Deep learning, and more specifically the pre-trained transformer paradigm, has substantially narrowed this gap; nevertheless, single-architecture solutions routinely leave accuracy on the table when applied to tasks that demand both rich contextual encoding and explicit sequential reasoning. This paper presents a cohesive, end-to-end AI- powered prediction framework that fuses BERT- derived contextual representations with a two-layer bidirectional LSTM (BiLSTM) classification head augmented by an additive attention mechanism. The system is designed as a modular pipeline: text acquisition and normalisation, augmentation-based imbalance handling, deep encoding, sequential modelling, and post-hoc probability calibration are treated as independent, replaceable stages. Experimental evaluation across three publicly available benchmark datasets — the LIAR fake news corpus, Stanford Sentiment Treebank v2, and a health-claim verification collection — confirms that the hybrid BERT-BiLSTM-Attention architecture outperforms five competitive baselines on macro-averaged F1 and area under the ROC curve. Ablation experiments quantify the individual contributions of the attention layer, recurrent head, augmentation strategy, and temperature scaling. A discussion of deployment trade-offs addresses inference latency, continual adaptation, and algorithmic fairness..
Do vision--language models (VLMs) develop more human-like sensitivity to linguistic concreteness than text-only large language models (LLMs) when both are evaluated with text-only prompts? We study this question with a controlled comparison between matched Llama text backbones and their Llama Vision counterparts across multiple model scales, treating multimodal pretraining as an ablation on perceptual grounding rather than access to images at inference. We measure concreteness effects at three complementary levels: (i) output behavior, by relating question-level concreteness to QA accuracy; (ii) embedding geometry, by testing whether representations organize along a concreteness axis; and (iii) attention dynamics, by quantifying context reliance via attention-entropy measures. In addition, we elicit token-level concreteness ratings from models and evaluate alignment to human norm distributions, testing whether multimodal training yields more human-consistent judgments. Across benchmarks and scales, VLMs show larger gains on more concrete inputs, exhibit clearer concreteness-structured representations, produce ratings that better match human norms, and display systematically different attention patterns consistent with increased grounding.
The Arabic neologism Aranjiyya (عَرَنْجِيَّة) is a portmanteau of ʿArabiyya (Arabic) and Inkliziyya / Faranjiyya (English/foreign), used by contemporary Arabic editors and stylists to describe Arabic prose that retains Arabic vocabulary while importing the syntactic, stylistic, semantic, or lexical structures of English. The phenomenon is pervasive in translated news, press releases, technical writing, and digital media, and is a recurrent target of prescriptive Arabic-style guides. Despite its prominence, Aranjiyya has had almost no presence in computational Arabic resources: existing treebanks and error corpora target orthographic, morphological, or syntactic well-formedness but do not isolate contact-induced patterns whose surface forms are grammatical but whose underlying templates are English. AranjiyyaCorpus was constructed to fill this gap, with three motivating use cases: training a span-level Aranjiyya detector for editors and translation post-editors; producing evaluation data for whether large language models actually generate idiomatic Arabic; and supporting linguistic study of contact-induced change in modern Arabic, with sufficient category granularity to distinguish syntactic, stylistic, and semantic phenomena and to track them across genres.
BACKGROUND: Atypical reward responsiveness is crucial for the onset and maintenance of non-suicidal self-injury (NSSI). There is no consensus whether NSSI is associated with increased or decreased reward responsiveness, due to different reward modalities, stages of reward processing examined, and the confounding effects of psychiatric comorbidities. This study aims to investigate reward responsiveness across different reward modalities—monetary and emotional—as well as distinct phases of reward processing, namely anticipatory and consummatory stages, among adolescents with NSSI. METHODS: The monetary and affective incentive delay (MID & AID) tasks were utilized to compare anticipatory and consummatory emotions as indicated by self-reported valence ratings between adolescents with NSSI (n = 103) and controls (n = 94) via ANOVA. The NSSI group was further separated into the NSSI + Suicidal Attempt (SA) and NSSI-only subgroups to examine whether NSSI co-occurring with suicidal behaviors resulted in more severe reward abnormalities. Sensitivity analyses using ANCOVA and hierarchical regression controlled for anhedonia symptoms to test the specificity of associations with NSSI. RESULTS: Adolescents with NSSI reported lower positive emotions when anticipating rewards, and experienced less pleasure after winning rewards or avoiding punishment. The NSSI + SA subgroup, compared with NSSI-only adolescents, had less consummatory pleasure in the MID task. However, blunted reward responsiveness in NSSI was no longer significant after controlling for anhedonia symptoms. CONCLUSIONS: NSSI is correlated with blunted reward responsiveness, primarily confounded by depressive symptoms (especially anhedonia). Co-occurring NSSI and SA is a more severe form of self-harm.
Cross-modal perception, the integration of information from multiple senses, plays a critical role in shaping emotional experiences. This study examines the interactions between visual and olfactory stimuli and their effects on emotional responses, a topic rarely addressed in prior research. Experiments employed five distinct visual stimulation methods that were combined with olfactory stimuli. Participants' emotional responses were assessed via surveys and electroencephalography (EEG) signal analysis. The study varied the color and movement direction of augmented particles to investigate their impact on EEG signals and emotional states. The findings demonstrated significant differences in emotional state classification under the influence of visual-olfactory interactions. Specifically, with backward-moving particles with matching colors (M4), classification accuracy was comparable to that of unimodal olfactory conditions (M1). Other visual stimuli generally caused confusion in classifying emotional responses. The increased valence ratings for pleasant aromas across all visual conditions did not consistently align with EEG-based classification results, suggesting that visual stimuli may introduce complexities into neural signals. These results highlight the intricate dynamics of multisensory interactions, emphasizing the role of visual stimuli in modulating emotional responses. The findings also suggest the potential of visual-olfactory interactions in developing augmented reality (AR) systems. By aligning visual and olfactory cues, AR environments can enhance the user experience and create immersive emotional landscapes, leading to applications for mood modulation and stress relief. This study underscores the relevance of multisensory integration in advancing emotion analysis and affective computing.
PURPOSE: Under a noisy environment such as a cocktail party, emotional signals play a crucial role in helping listeners unmask target speech. However, it remains unclear how emotional features carried in a speaker's vocal timbre shape neural processing over time. This study aimed to characterize the temporal neural dynamics of learned emotion with a speaker's voice in complex listening conditions. METHOD: We employed an emotional learning paradigm in a speech-on-speech context, pairing two different target speakers with either angry or neutral facial expressions. Electroencephalogram data were recorded from healthy participants, and multivariate pattern analysis combined with representational similarity analysis was used to track the temporal unfolding of learned emotion linked to the target speaker's voice. RESULTS: We observed early neural signatures of emotional processing between 150 and 180 ms after stimulus onset, occurring nearly simultaneously with the decoding of speaker identity. Importantly, brain-behavior analysis revealed that subjective emotional valence ratings could be decoded from neural signals as early as 94 ms. These findings suggest that vocal emotion can be processed rapidly and in a way relatively independent to the process of low-level acoustic cues. CONCLUSION: Our study provides evidence that acquired emotional associations with a speaker's voice can shape early-stage neural dynamics during speech processing under challenging listening conditions. SUPPLEMENTAL MATERIAL: https://doi.org/10.23641/asha.31842814.
An open, end-to-end PROIEL-style treebank of the entire Greek language from Homeric and Archaic texts (8th c. BC) through Classical, Koine, Late Antique, Byzantine, Late Byzantine, Early Modern, and Modern Greek, with cross-lingual alignment to Latin (Vulgate), Gothic (Wulfila), Old Church Slavonic (Marianus), and Classical Armenian via the verse-level cross-lingual alignment of the New Testament. Includes Stanza PROIEL-trained dependency parses, LaBSE sentence-level + mBERT word-level alignment, and ASJP/LingPy phonetic cognate scoring. Part of the Athens Digital Glossa Chronos Research Network (AthDGC) and the CVL-CDSAML project (A Corpus-based Valency Lexicon for a Contrastive and Diachronic Study of Ancient and Medieval Languages), funded by the Hellenic Foundation for Research and Innovation (HFRI) under the 3rd Call for HFRI Research Projects to support Post-Doctoral Researchers, Project No. 20577, with support from the Greece 2.0 National Recovery and Resilience Plan. Hosted at the National and Kapodistrian University of Athens (EKPA), Division of Language-Linguistics, Department of English Language and Literature, School of Philosophy.
Background: Word identification in noise is crucial for effective communication in everyday environments. For children, the ability to identify words in noisy conditions directly impacts language development, learning, and social interaction. This study aimed to develop and standardize Hindi word identification in noise test for children (HWINT-C) and evaluate its performance among school aged typically developing children across varying SNR and word length. Methods: The study included forty two participants which were further subdivided into subgroup one consisting of 22 typically developing children aged 6-7.11 years (SGI) and 20 typically developing children aged 8-10 years in subgroup two (SGII). Development of Hindi word identification in noise test for children (HWINT-C) involved multi-step processes including selection of words, familiarity rating, and content validation, internal consistency and test-retest reliability. The test included bisyllabic and monosyllabic words recorded by a native Hindi female speaker presented in eight-talker babble at +5 dB and +7 dB SNR administered dioticallyat 65 dB SPL. Results: The developed HWINT-C in this study demonstrated to have high internal consistency and test-retest reliability. Typically developing children in the older group (SGII) significantly outperformed the younger group (SGI), Additionally performance improved with increasing signal-to-noise ratio (SNR) in both SGI and SGII, but no significant differences were found across word lengths. Conclusions: The HWINT-C test is a reliable and valid tool for assessing word-in-noise perception in children. Age-related trend was observed, where performance improved with age, Similar findings were observed with increase in SNR, emphasizing need of favorable conditions in younger population.
The subject of the article is the historical dynamics of the formation and use of feminitives in Russian and English and their role as markers and constructors in the verbalization of gender roles. Using the example of two language groups, the sources of the emergence of feminitives, the stages of normalization and unconscious use of them, as well as current trends in the processing of feminitive vocabulary in the context of gender neutrality, inclusivity and social linguistic norms are considered. The article examines the historical background of the formation and use of feminitives in Russian and English, as well as their role as markers and constructs reflecting and shaping gender roles in verbalization. The object of the study is femininity, used in Russian and English in various contexts. The article uses the method of comparative analysis of the features of the formation of feminitives in Russian and English. The research methodology consists of historical and linguistic analysis of language corpora, comparison of diachronic and synchronous data, study of academic, social and everyday texts and gender-marked vocabulary. The relevance of the topic is due to the verbal conflicts between traditional gender norms and movements for gender equality, which are reflected precisely in the vocabulary of professions, roles and identity. The novelty lies in conducting a comparative analysis that allows us to identify how gender roles are reflected and constructed in each culture through specific designations of women's professions, positions and social statuses. Linguistic comparative analysis makes it possible to identify how gender stereotypes are verbalized in each culture through specific designations of women's professions, positions and social statuses. This approach makes it possible to trace the differences and similarities in the processes of feminization of vocabulary, as well as to understand how the use of feminitives contributes to the formation and understanding of gender relations in Russian and English.