Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
18625 papers
This study reviews the English language test of Singapore’s Primary School Leaving Examination, a high-stakes national assessment taken annually by nearly all primary six students for secondary school placement. Given the test’s importance in shaping students’ academic pathways and recent format changes, it is crucial to evaluate its validity, specifically its ability to provide accurate and fair assessments of students’ English language proficiency and academic readiness. The review outlines the test’s educational and policy context, followed by a description of the latest formats for both the English language and foundation English language versions. The analysis focuses on core dimensions of test validity, including content representativeness, construct validity, criterion-related validity (concurrent and predictive), and reliability (inter-rater reliability and internal consistency). Drawing on official documents and limited empirical studies, the review finds moderate improvements in content representativeness and construct validity. However, both longstanding and emerging concerns (e.g., the exclusion of local linguistic norms and genre scope) indicate that key limitations remain. While predictive validity, inter-rater reliability, and internal consistency appear supported, empirical research remains sparse across all reviewed test qualities, particularly in concurrent validity. The review integrates identified research gaps and proposes inquiry directions to inform future test development and policy adaptation. Strengthening the evidence base is essential for ensuring a valid, reliable, and equitable assessment system in Singapore’s primary education landscape.
Word Sense Disambiguation (WSD) is a fundamental task in Natural Language Processing (NLP), addressing the challenge of identifying correct word meanings in context. This task is particularly complex for morphologically rich and resource-limited languages like Hindi, which exhibit significant lexical ambiguity compounded by limited availability of annotated corpora. To address these challenges, we propose a supervised approach combining the multilingual BERT model (mBERT) with Hindi WordNet as a structured lexical resource. Using few-shot learning, we fine-tune mBERT on a dataset constructed from Hindi WordNet to disambiguate contextually ambiguous words across four parts of speech (POS): nouns, verbs, adjectives, and adverbs. Experiments on standard Hindi WSD benchmarks demonstrate that our method significantly outperforms traditional rule-based and embedding-based approaches, achieving 96.48% accuracy—an approximate 3% improvement over the strongest baseline. These results validate the effectiveness of integrating contextualized embeddings from pre-trained language models with structured lexical databases, highlighting the promise of hybrid techniques for advancing WSD in low-resource languages and providing a framework applicable to other morphologically complex languages with similar resource constraints.
This paper examines how social media discourse affects the learning of the English language on the tertiary level in Lahore and how informal online communication influences the academic English of students in Lahore. The qualitative design was employed to gather data on the basis of semi-structured interviews with English language learners and teachers. Braun and Clarke’s (2006) framework was used to conduct thematic analysis on the transcribed data. The results indicate that though social media provides appropriate exposure to vocabulary, pronunciation, and language use in real life, the casualness of linguistic norms has a very strong impact on the students’ academic writing and their communicative accuracy. The participants claimed to use short forms, abbreviations, slang and mixed-language texting on a regular basis, and these transferred to essays and presentations. Other themes were distraction and a lack of studying discipline, the inability to stick to the formal register, and the misunderstanding of words acquired in unconfirmed online situations. Teachers also reported on the same lines, as they observed poor writing standards in academic writing and increased dependence on social-media-driven language patterns. Pedagogical mechanisms to counteract these effects were also determined in the study with the focus being on register awareness, purposeful digital task integration, and curriculum modernization. On the whole, the study has determined that the power of social media is twofold, both positive when moderated and negative when uncontrolled and recommends that informed teaching and learning methods are needed to help students balance between informal online communication and formal academic language.
The Book of Dede Qorqud, a key work of Turkic oral literature, provides valuable insights into the socio-cultural, linguistic, and psychological world of the Oghuz Turks. The forms of address between men and women reflect gender roles, interpersonal relationships, and communication strategies. Women's speech often conveys politeness and emotional appeal, while men's emphasizes authority and social hierarchy. Linguistically, address terms reveal intimacy, respect, and power dynamics through lexical choices and formal or informal expressions. Psycholinguistically, women's nurturing and persuasive speech highlights their role in maintaining harmony, while men's assertive forms reflect their expected authority. Overall, these address strategies illustrate cultural expression, social regulation, and gender dynamics in Oghuz society, showing how language shaped identities and relationships within the framework of tradition and norms.
Large language models (LLMs) have reignited debate about whether machines without minds or intentions can genuinely participate in linguistic practice. Critics portray them as ‘stochastic parrots’ that manipulate form without meaning, whereas defenders emphasize their impressive functional capacities. This paper argues that these disputes conflate distinct dimensions of meaning and agency. I extend Huw Price’s distinction between i-representation and e-representation (roughly, inferential versus environment-tracking types of representation) by differentiating physical e-representation—such as a fuel gauge, grounded in causal coupling—from symbolic e-representation, exemplified in language and mediated by agents. This refinement clarifies what is at issue: LLMs clearly display i-representational competence through their participation in inferentially structured discourse. Whether their outputs possess symbolic e-representational content, however, is contested and framework-relative. It depends on whether agent-mediated uptake is taken to suffice, or whether additional grounding conditions—such as intentions, causal connections, or proper functions—are required. I further distinguish norm-sensitivity—the capacity to track and adapt to linguistic norms, which grounds their i-representational competence—from norm-responsibility, the reflexive capacity to own commitments and bear accountability. Technical analysis of LLM architectures shows that they exhibit advanced norm-sensitivity through statistical learning but entirely lack norm-responsibility. LLMs thus occupy a distinctive position: they are genuine functional participants in linguistic practices, yet fall short of the reflexive agency characteristic of responsible speakers.
Abstract This paper presents a novel framework for modeling role and task allocation in cooperative wheeled soccer robot systems by leveraging latent knowledge extracted from past collaborative interactions. Inspired by recent advances in heterogeneous multi-robot collaboration, the proposed method encodes a soccer team as a set of Multidimensional Relational Structures (MDRSs), capturing both temporal and spatial relations among robot roles, actions, and stimuli. A structured dataset, termed the Soccer Robot Collaboration Treebank (SRCT), is introduced to represent play-by-play histories of robot behaviors, parsed through a formal grammar to support structured learning. Probabilistic modeling and Non-Negative Tensor Decomposition (NTD) are applied to the resulting tensors, enabling robust inference and latent knowledge estimation even in scenarios with sparse data or communication loss. Simulated experiments using a team of wheeled soccer robots in the Webots environment demonstrate the system’s ability to dynamically reassign roles, reason over incomplete histories, and predict collaborative behaviors such as passing, defending, or role-switching. The results show that the proposed framework enhances both strategic flexibility and robustness, providing a foundation for real-time decision-making in robotic soccer under uncertainty.
The article is devoted to studying socially determined linguistic processes, which are traditionally associated with the broad problem of social variation of communication (discourse) and linguistic variability of the German language. It presents the results of a study conducted within the framework of cognitive sociolinguistics - the linguistics of social meanings. The author observes continuity in the development of scientific thought, explores the problem of linguistic variability in two modes - theoretical and applied. The subject of the research is the terminological apparatus and specific linguistic facts that are based on the cognitive, communicative and social functions of language, that is, to be the reality of the thought of individual and/or collective knowledge of representatives of a certain society and the means of their communication. The purpose of the article is to analyze theoretical propositions, linguistic terms and existing specific language forms that convey social meanings, marked by a social feature in their terminological interpretations with a focus on the linguistic picture of the Germany new lands. The novelty of the research lies in cognitive-semantic analysis, systematization and modeling this phenomenon in the social aspect. As a result, the author comes to her own conclusions, which lead to understanding and rethinking the old views on the problems of dialectology in the context of modern realities of the German language society and the data of modern linguistics. The methodology of the research and the description of its results are determined by the principle of interdependence of the three most important didactic and linguistic strata - the study of modern language from the standpoint of linguistic norms, the study of linguistic variability and the analysis of language change in the framework of its historical development. General scientific methods and special methods of cognitive linguistics are used to analyze theoretical material and linguistic facts, including explanatory description, interpretation, cognitive modeling, cognitive dominance, and focusing.
Pre-trained Language Models (PLM) have enabled a cost-effective approach to handling various downstream applications via Parameter-Efficient-Fine-Tuning (PEFT) techniques. In this context, service providers have introduced a popular fine-tuning-based product service known as Model-as-a-Service (MaaS). This service offers users access to extensive PLMs and training resources. With MaaS, users can fine-tune, deploy, and utilize their customized models seamlessly, leveraging a one-stop platform that allows them to work with their private datasets efficiently. However, this service paradigm has recently been exposed to the possibility of leaking user private data. To this end, we identify the data privacy leakage risks in MaaS-based PEFT and propose a Split-and-Privatize (SAP) framework, mitigating the privacy leakage by integrating split learning and differential privacy into MaaS PEFT. Furthermore, we propose Contributing-Token-Identification (CTI), a novel method to balance model utility degradation and privacy leakage. As a result, the proposed framework is comprehensively evaluated, demonstrating a 65% improvement in empirical privacy with only a 1% degradation in model performance on the Stanford Sentiment Treebank dataset, outperforming existing state-of-the-art baselines.
This article explores the critical engagement of two academics who confront lived experiences with the institutional and tangible dimensions of linguistic barriers and discrimination in the Portuguese district of Faro. Centred on the challenges posed at the border, the study fits into the wider framework of mobilities between North Africa and southern Europe, and attempts to demonstrate how language, as a form of social practice, impacts access to employment, education and society at large. Portuguese emerges as a dual entity: an institutional barrier, a form of socio-spatial control that reinforces exclusion, illustrating exclusionary processes within hierarchies and structural violence; a transnational bridge that fosters belonging and, simultaneously, a battleground where identity and self-determination face the constraints imposed by the economic, social, and political order that impact Moroccan migration to southern Portugal. The research highlights the imbrications of the undervaluation of migrants' cultural knowledge, the ambivalence of linguistic identity within a globalized world, exacerbating social exclusion, and systemic discrimination of non-privileged migrants in a polarized region shaped by social and economic asymmetries and increasingly representative nationalisms. From a systemic justice perspective, this devaluation reinforces structural inequalities, marginalizing those who do not conform to dominant linguistic norms. The national languages’ role as linguistic and cultural gatekeeper exemplifies the intersection of identity construction and socio-political hierarchies in the context of mobilities. This study, grounded in a collaborative project blending autobiography and biographical research, employs qualitative methods, including biographic interviews, ethnographic observation, and critical human rights studies.
Mood, an individual’s emotional state, fundamentally shapes how the brain interprets sensory input by providing a continuous affective context for prediction and evaluation. In language processing, mood may bias the interpretation of emotionally valenced words, amplifying or dampening their perceived affect. Yet, the temporal dynamics of these mood-valence interactions remain poorly understood. To clarify inconsistent evidence on the timing and nature of mood-valence interactions, we examined how induced mood influences early stages of emotional word processing using EEG. Participants performed a valence-rating task for positive, negative, and neutral words in a baseline condition and following positive or negative mood induction. Event-related potentials were analysed across early processing windows (N1, P2, EPN) using cluster-based permutation statistics. Positive mood selectively attenuated N1 amplitudes for highly valenced words, consistent with reduced prediction error under mood-congruent expectations. Later components (P2, EPN) showed decreased amplitudes for both high and neutral valence, suggesting reduced model updating under mood-congruent expectations. Negative mood, in contrast, produced weaker and temporally delayed modulations. Behaviourally, participants responded more quickly to valenced words under induced mood conditions, supporting the neural findings. Interpreted within a predictive coding framework, these results support the theoretical view that mood functions as a hyperprior, tuning the precision of predictive models during language comprehension. Positive mood appears to enhance predictive flexibility and facilitate the processing of affectively congruent words, whereas induced negative mood reduces positive affect. Taken together, the findings highlight how affective states dynamically modulate early predictive mechanisms in emotional language processing.
The deployment of large language models (LLMs) across heterogeneous environments requires format-specific conversion, precision tuning, and consistent evaluation-tasks that are often fragmented across multiple tools. This work presents SOLO-Export, a unified command-line interface (CLI) framework for multi-format export and post-export benchmarking of causal LLMs. Having precision options for FP16 and INT8 where appropriate, the system supports the ONNX, TorchScript, Hugging Face, TensorFlow Lite, and TensorRT backends. Device-aware exports for both CPU and CUDA targets are made possible by a configuration-driven workflow that generates artifacts in a uniform directory structure. Each exported model is benchmarked using the Penn Treebank dataset by the integrated evaluation harness, which reports inference latency, token-level accuracy, and perplexity. According to experimental results, FP16 exports on GPU-oriented backends like TensorRT achieved up to 3.2 times lower latency than baseline FP32 models. On the top of that, with minimal impact on perplexity, storage size was reduced by more than 60% thanks to INT8 quantization. The combined approach reduces manual configuration overhead, speeds up deployment preparation, and ensures consistent performance insights across formats. This study demonstrates that a single, scalable pipeline can be effective.
Abstract Older adults consistently report higher emotional well-being despite some physical and mental declines with age. Some theorize this is due to differences in emotion regulation, however no conclusive evidence for age differences in emotion regulation strategies has emerged. Emotion regulation tactics, such as positive-approaching (enhancing the positivity of a situation) and negative-receding (reducing the negativity of a situation), hold promise for uncovering age differences. However, no studies to date have considered whether tactic vary in their physiological profiles. Thus, this study investigated 35 younger (M = 19.06 years, SD = 3.58; 17 women) and 42 older (M = 74.02 years, SD = 4.69; 23 women) participants who viewed emotionally-evocative videos and regulated emotions using positive-approaching and negative-receding techniques (with both regulation blocks compared to a neutral video block to control for baseline differences in psychophysiology). Multi-level linear models revealed significant Age x Sex x Tactic interactions, such that positive-approaching tactics (vs negative-receding) were associated with lower arousal ratings for younger women and older men (but not younger men or older women). While there was no tactic difference in self-reported arousal for older women, Respiratory Sinus Arrythmia (a measure of heart activity associated with calm parasympathetic states) was higher during positive-approaching for them. Interestingly, older men had higher RSA during negative-approaching tactics as compared to positive-approaching. These findings suggest emotion regulation tactics have age and sex differential impacts on self-reported and physiological arousal. Future work may disentangle physiological activation and self-reported behavior by also considering the role of interoceptive awareness.
When stimuli are retained in visual working memory (VWM), external stimuli which overlap this representation capture attention when performing a visual task. It has not been determined whether this mechanism can partly account for attentional capture by categories of real-world affective stimuli. Across five dual-task visual search and VWM change detection experiments (4/5 pre-registered; total N = 119) participants had to detect the change in either positive (kitten) or threat-related (spider) animal exemplars, whilst performing an intervening visual search task with peripheral distractors from these affective categories. Affective stimulus associations were confirmed by self-reported arousal and valence ratings in all samples, and confirmed in an independent sample (n = 82). It was hypothesised that threat-related and positive distractors would capture attention more, versus a neutral (bird or no distractor) baseline, when matching the contents of VWM. Experiments 1 - 3, however, found no evidence of increased capture by VWM-matching affective stimuli, though there was cumulative evidence of goal-independent capture by threat-related distractors. When, however, the trial structure became unpredictable, requiring constant preparation for the VWM task response (Experiment 4), or advanced action preparation to the VWM task was enabled (Experiment 5), then VWM-matching threat-related distractors caused greater attentional capture. This VWM-driven capture, however, was not found for positive distractors in any experiments. The results probe the boundary conditions when VWM contents drive attentional capture by entirely task-irrelevant affective categories, and suggests that background memory representations may not influence attention unconditionally, and instead may depend partly on their current prioritisation.
Math anxiety poses significant challenges for university psychology students, affecting their career choices and overall well-being. This study employs a framework based on behavioural forma mentis networks (i.e. cognitive models that map how individuals structure their associative knowledge and emotional perceptions of concepts) to explore individual and group differences in the perception and association of concepts related to math and anxiety. We conducted 4 experiments involving psychology undergraduates from 2 samples (n1 = 70, n2 = 57) compared against GPT-simulated students (GPT-3.5: n2 = 300; GPT-4o: n4 = 300). Experiments 1, 2, and 3 employ individual-level network features to predict psychometric scores for math anxiety and its facets (observational, social and evaluational) from the Math Anxiety Scale. Experiment 4 focuses on group-level perceptions extracted from human students, GPT-3.5 and GPT-4o's networks. Results indicate that, in students, positive valence ratings and higher network degree for "anxiety", together with negative ratings for "math", can predict higher total and evaluative math anxiety. In contrast, these models do not work on GPT-based data because of differences in simulated networks and psychometric scores compared to humans. These results were also reconciled with differences found in the ways that high/low subgroups of simulated and real students framed semantically and emotionally STEM concepts. High math-anxiety students collectively framed "anxiety" in an emotionally polarising way, absent in the negative perception of low math-anxiety students. "Science" was rated positively, but contrasted against the negative perception of "math". These findings underscore the importance of understanding concept perception and associations in managing students' math anxiety.
Recent studies demonstrate that visual working memory capacity is greater for real-world objects compared to simple features like colors and scrambled objects. This led to the proposal that conceptual meaning plays a critical role in structuring visual working memory (Chung, Brady, & Störmer, 2024). However, one challenge in comparing memory performance across stimulus sets is that they vary not only in conceptual meaning but also in perceptual similarity. Thus, some of the working memory benefits for real-world objects may arise from these perceptual differences – for example whether the visual system interprets inputs as objects or not. Here, we provide a strong test of this by using novel objects generated by generative adversarial networks designed to resemble real objects (Cooper et al., 2023), and compare memory performance across novel and familiar real-world objects. Across experiments, participants remembered sets of four objects drawn from one of four stimulus types: familiar objects, novel objects, scrambled familiar objects or scrambled novel objects. After a short delay, they completed a two-alternative-forced-choice task, selecting between a target and a foil object. Results revealed enhanced working memory performance only for familiar objects, with no differences among the other conditions. Importantly, convolutional neural networks analyses confirmed comparable perceptual similarities between familiar and novel objects relative to scrambled stimuli. Thus, although novel objects closely resembled familiar objects, they did not enhance memory performance. Further correlation analyses revealed that subjective familiarity ratings are correlated with memory performance for familiar objects, while low-level features like colorfulness are correlated with memory performance for novel objects, suggesting that visual memory relies on different aspects to best remember each stimulus type. Overall, these results demonstrate that “object-ness” alone is insufficient to enhance visual working memory. Instead, familiarity and conceptual knowledge are critical in improving working memory performance.
“Pictures are worth a thousand words," yet most platforms like Yelp, Google Maps, Instagram, Walmart, and Amazon require users to provide text, ratings, and images. Images often capture a user's intent, and the features within the images typically correlate with that intent. In this paper, we extract various features from images (such as edge distribution, color distribution, text within the image, focus, etc.) and compare simple vs. complex models to predict the ratings associated with these images. We find that features such as brightness and contrast significantly explain the rating at image-level, and models such as random forest and logistic regression provide a 0.84 F-1 score when predicting the rating. In the era of generative AI, we anticipate that sharing an image will allow platforms to auto-generate user intent and image ratings, thereby simplifying the dissemination of information.
Previous research regarding verb production deficits in Alzheimer’s disease (AD) primarily concentrated on either the quantity of verbs (inflections) or verb-related semantic units, with little consideration given to verb production within syntactic contexts, i.e., verb collocations. This study explored verb collocations in the connected speech of Chinese AD patients within the framework of dependency syntax. The findings include: (1) The frequency distribution of verb collocation patterns in AD follows the Mixed-Poisson function similar to that in the healthy control elderly (HCE) and healthy control young (HCY) groups, but it differs in the use of low- and high-collocation patterns; (2) In the static aspect, the AD patients exhibit the lowest overall mean collocation pattern (MCP) among the three treebanks, followed by the HCE group. In the dynamic aspect, the MCP and sentence length in the three groups show a similar synergistic relation, but differences exist in the quadratic regression parameters; (3) Based on the probabilistic distribution of verb-governed dependencies, the AD patients exhibit the lowest syntactic proficiency, followed by the HCE group. The differences between the AD patients and the HCE group confirm the presence of verb production deficits and a decline in syntactic proficiency in AD. Although the HCE group also shows mild language deterioration compared to the HCY group, the extent of these changes is considerably smaller than that observed in the AD patients. These findings suggest that while aging may contribute to a partial decline in language abilities, AD markedly exacerbates and accelerates this deterioration process, following a pathological trajectory distinct from normal aging.
Learner Handover (LH) involves sharing information about learners between faculty supervisors, aligning with a growth mindset. Previous studies, however, demonstrate LH can bias subsequent ratings. Most of these studies collect ratings after a single encounter but faculty often have multiple interactions with learners potentially mitigating LH-related bias. This study explored if LH influences faculty ratings, entrustment decisions and feedback after observing several encounters of the same learner. Internal medicine faculty (n = 57) from five medical schools were randomly assigned to one of three study groups. Each group received either positive, negative or no LH prior to watching five simulated resident-patient encounter videos of the same white male resident. Participants rated each video using an entrustment scale, the Mini-CEX and provided written feedback. Feedback was assigned a valence score (-3 to + 3). There were no statistically significant differences between the mean ratings across the LH conditions (positive, control, negative) for entrustment [3.42, 3.26, 3.62], Mini-CEX [6.00, 5.90, 6.28] or feedback valence ratings [-0.34, -0.99, -0.74]. In the post-study questionnaire, most raters reported the LH had minimal effect on their decisions. Only 29% of raters guessed the true purpose of the study. Unlike previous studies, LH had no effect on ratings, entrustment decisions, or feedback after one encounter, nor over subsequent encounters with the same resident. These findings suggest LH's influence may vary and highlight the need for replication under different conditions, including diverse genders and equity-deserving groups, to identify factors that contribute to or mitigate bias.
Large Language Models (LLMs) employ deep learning algorithms to generalize patterns in data. Applying these LLMs to classification tasks can reduce the required labor and time. The research aims to fine-tune the LLM Llama 3.1 to correctly identify whether a chosen text message exhibits a positive or negative emotion. The goal of this procedure is to apply the fine-tuned LLM to large databases of text messages and locate users whose recent texts contain a large proportion of negative samples. This way, I can alert the users and direct them to help very early on. I chose the Stanford Sentiment Treebank v2 (SST-2) dataset. It mimics the emotional polarity of real texts with its even positive-negative sample distribution and its contextless format. I used the Unsloth framework and LoRa to significantly reduce the resources required during the fine-tuning process. I tested the model by taking SST-2’s train split and inputting them individually into the trained model. Using this method, I found the Llama model to be highly accurate, with an accuracy of 94.8%. Interestingly, it had a high average Binary Cross-Entropy (BCE) Loss of 0.782 but achieved high accuracy. The testing against other models shows that the BCE Loss for sentiment analysis is not correlated to the actual accuracy of the model. From the results, I determined Llama 3.1 was the most suitable LLM for the sentiment analysis of large text databases.
PURPOSE: This study interrogates the intercultural experiences of African international students (AIS) in a US Midwestern University. With a focus on West African students, the study explores how students confront and overcome linguistic, cultural, and systemic barriers as they create new dynamic spaces which are neither American nor African, but a mixture of both. SUBJECTS: 12 West African students. METHODS: I utilized ethnographic observations and in-depth interviews to better understand how the AIS engage in ongoing identity negotiation through language, food, cultural expressions, and technology. I conducted 40 hours of participant observation research over a 5-month period and interviewed 12 West African students. OUTCOME: Drawing on the theoretical constructs of hybridity and language ideology, the findings reveal how AIS negotiate identity by using traditional African cultural traits, rooted in what I term Africanism—with new cultural influences, generating unique, fluid and constantly-evolving hybrid identities. Language emerges as a powerful site of hybridity, where students shift between linguistic norms to balance intelligibility and cultural authenticity. These hybrid identities emerge through active, creative processes. IMPACT: The findings offer important implications for higher education, emphasizing the need for more culturally responsive support systems that recognize the distinct experiences of African students. Additionally, the study contributes to the fields of communication and migration studies by advancing a nuanced understanding of identity formation in transnational contexts. By foregrounding the voices of African students, this research challenges monolithic representations of international students and promotes intercultural dialogue in fostering inclusive communities and enriching educational environments.
The research paper is devoted to a comprehensive consideration of the mediative function of language in traditional Kazakh culture. The main purpose of the work is to identify the specifics of the use of language in the historical and cultural practice of Kazakh society as a means of coordinating interests, settling disputes and harmonizing social relations. In accordance with this goal, the article defines the following tasks: cultural and social foundations of linguistic mediation in traditional Kazakh society; description of linguistic structures and pragmatic strategies of institutions that ensure dispute resolution through speech; identification of semantic features of linguistic norms aimed at maintaining social harmony in national culture. In the course of solving these tasks, the oral oratorical heritage of the Kazakhs is analyzed from linguistic and pragmatic positions. The conducted research allowed us to establish that the mediative function of language in traditional Kazakh culture goes beyond simple communication. It forms the basis of mechanisms for maintaining social harmony, regulating the moral code of the community and the peaceful settlement of conflict situations. It is also proved that the speech culture of leaders and bi-speakers contributed to the formation of a specific national model of mediation, and its language strategies (forms of etiquette, indirect ways of expression, metaphorical and symbolic structures and pragmatic means of mitigation) are consonant with modern theories of mediation. The scientific significance of the study lies in the fact that the phenomenon of mediation in Kazakh culture is being systematically examined for the first time in a linguistic and pragmatic perspective, which makes it possible to identify the contribution of traditional speech experience to the development of the general theory of mediation. The practical value of the work is determined by the possibility of using the results obtained in mediation training programs, in the development of ethno-cultural models of negotiation practices, as well as in projects aimed at updating the culture of Kazakh oral speech.
The article examines the case forms of nouns in Ulas Samchuk’s novel «Maria» from the perspective of modern linguistic norms. In recent decades, we have observed significant variability in the use of case endings of nouns in the modern Ukrainian language, due to the restoration or activation of many ancient forms. The novel «Maria» by the prominent Ukrainian writer Ulas Samchuk, written in 1933 was chosen exactly for confirmation of the continuity of the grammatical tradition. The proposed study examines the case forms of the dative, accusative, locative, and vocative cases. It was found that the most of the case forms of nouns recorded in the novel are normative even today: the ending of the dative case of masculine nouns -ові, -еві (-єві), which actively displacе the ending -у(-ю), are increasingly penetrating the system of neuter nouns; a significant spread of genitive case forms in the function of the accusative in nouns – names of non-beings; variant forms of the local case (masculine and neuter nouns) in -у and -і in constructions with the preposition по; consistent use of vocative case forms in appeals; alternation of consonants г, к, х in the local and vocative cases. Some case forms of nouns observed in the analyzed novel are used less frequently in modern linguistic practice, while others are qualified as dialectal. It is concluded that the case forms of nouns used in Ulas Samchuk's novel «Maria» reflect the grammatical norms of the Ukrainian language of the first half of the 20th century, many of which were artificially brought closer to the norms of the Russian language in the following decades or relegated to the periphery of the language system due to the socio-political situation. «Ukrainian Spelling» 2019, by bringing back to life some features of the spelling of 1928, renewed the Ukrainian orthographic tradition, which is clearly evidenced by the work of Ulas Samchuk.
The ability to complete sentences and guess words is an essential component of NLP. Virtual assistants and advanced technology that dictate text to humans both gain from them. Because of their ambiguity, reliance on long-term events, and lack of information, humans are currently unable to complete these jobs. Common systems use RNNs and transformer topologies; however, these models may struggle to understand and identify uncommon words. Such limitations may hinder their productivity and make it more difficult for them to put their knowledge into practice. To overcome these challenges, this work employs federated learning for word guessing and N-gram modelling. To identify linguistic trends that manifest in many locations, the federated N-grams approach makes use of a large number of computers. Both bias and missing data can be reduced in this way. The Penn Treebank and AS WIKI-text are two of the basic datasets that we use to train and evaluate our model. Using optimisation techniques, we can understand the situation, handle unusual words, and account for bias in our guesses. The results of this study show that Fed-based learning has the potential to revolutionise language modelling by removing control mechanisms from AI while simultaneously protecting users' personal information. With the use of the N-grams method, predictions regarding the next word could be more precise and contextually aware.
The article describes intra- and extra-linguistic factors that influence the stability of the onymic space of the Ukrainian language. Intra-linguistic factors are the creation of a certain proper name according to its inherent derivational model; the correspondence of a particular onymic to the formed linguistic norm. Extra-linguistic factors are the level of linguistic, spiritual and political culture of society and its national consciousness. This influence is most pronounced on two classes of onymic vocabulary – anthroponyms and toponyms, as well as on toponymic derivatives – the names of the inhabitants of the corresponding settlement (katoikonyms) and derived adjectives (adjectonyms). It is specially noted of the three-component anthroponymic formula (first name, patronymic, surname) in modern Ukrainian official speech; the elimination of variation in the declension of Ukrainian surnames of the masculine gender with the possessive suffix –ів. Іt is shown that main means of creating katoikonyms in the Ukrainian language are the suffixes –ц-і (plural), –ець, –к-а (singular). Possible functioning of parallel catoikonymic forms with suffixes –ц-і (-ець, –к-а) й –ан-и / –ян-и (-ан-ин, –ан-к-а / –ян-ин, –ян-к-а). The paper deals with problem of establishing derivational models of adjectonyms and their normalization. The author considers that it is necessary to carefully study the historical patterns of the creation and use of adjectonyms in the Ukrainian language, as well as to take into account local traditions. The destructive impact on the Ukrainian oikonymic space of numerous unmotivated ideological renamings of the Soviet era and artificial formations with a Russian-language structure is analyzed. The author shows that it is necessary to cleanse the Ukrainian oikonomika and urbanonymika of such names. Keywords: adjectonyms, anthroponyms, katoikonyms, oikonyms, onymous space, urbanonyms, artificial names (renaming).
With the rise of large language models, service providers offer language models as a service, enabling users to fine-tune customized models via uploaded private datasets. However, this raises concerns about sensitive data leakage. Prior methods, relying on differential privacy within device-cloud collaboration frameworks, struggle to balance privacy and utility, exposing users to inference attacks or degrading fine-tuning performance. To address this, we propose PrivTune, an efficient and privacy-preserving fine-tuning framework via Split Learning (SL). The key idea of PrivTune is to inject crafted noise into token representations from the SL bottom model, making each token resemble the $n$-hop indirect neighbors. PrivTune formulates this as an optimization problem to compute the optimal noise vector, aligning with defense-utility goals. On this basis, it then adjusts the parameters (i.e., mean) of the $d_χ$-Privacy noise distribution to align with the optimization direction and scales the noise according to token importance to minimize distortion. Experiments on five datasets (covering both classification and generation tasks) against three embedding inversion and three attribute inference attacks show that, using RoBERTa on the Stanford Sentiment Treebank dataset, PrivTune reduces the attack success rate to 10% with only a 3.33% drop in utility performance, outperforming state-of-the-art baselines.
Исследование посвящено совершенствованию метрик релевантности эмоционального контекста в системах извлечения контекста для Retrieval-Augmented Generation. Современные подходы фокусируются на семантическом сходстве текстов, пропуская эмоциональную составляющую, когерентность, что критично для задач наиболее точного поиска в эмоционально окрашенных данных. Предложена композитная метрика, объединяющая семантическое сходство векторных эмбеддингов и эмоциональную когерентность, оцениваемую как косинусное сходство текстов. Описана методология вычисления компонент метрики и их взвешивания, на основе которой проведен ряд экспериментов с целью выяснения способности предложенной метрики находить текст более точно, чем модель, основывающаяся на поиске только семантической составляющей. Оценка эффективности проведена на сэмплированном подмножестве датасета Stanford Sentiment Treebank 2 (SST-2). Композитная метрика продемонстрировала более высокую эффективность по метрикам Precision@5, Recall@5, NDCG@5 и MAP по сравнению с методами поиска, основанными исключительно на семантической составляющей. Проведенные эксперименты показали значительный прирост метрик для позитивных и негативных запросов. Проведен дальнейший анализ влияния коэффициентов предложенной метрики. При рассмотрении полученных результатов было установлено, что учет эмоциональной когерентности необходим для достижения оптимальной производительности метрики.
Abstract The legal language has been extensively studied due to its profound and far-reaching impact on society. Among the various research areas, the complexity of legal language has received particular attention due to its implications for legal practice and societal access to justice. Relevant studies seek to identify the factors contributing to the difficulty of legal language and explore potential methods of simplification. However, despite extensive qualitative research, a comprehensive, quantitative analysis of the essential linguistic features distinguishing legal language from everyday language remains underexplored. The present study fills this gap and quantitatively examines the key features of German legal language by comparing it to journalistic language as a benchmark for everyday discourse, providing a rigorous, corpus-based perspective on the lexical, syntactic, and typological features of legal texts. Using data from the HDT-UD and DGT-UD treebanks, the study analyzes 18 quantitative indicators across three linguistic domains – lexicon, syntax, and word-order typology. The results indicate that legal language is characterized by a limited vocabulary, a higher frequency of long words and dependent clauses, increased syntactic and structural complexity, and a predominance of SV and OV word order patterns. By providing a detailed comparison of legal and journalistic registers, this study advances the understanding of legal language through objective, empirical analysis. The findings have practical implications for legal communication, suggesting that greater attention to lexical simplification and syntactic clarity could improve accessibility and comprehension.
The Philippines, having over 170 languages, faces language barriers and diminished language fluency due to heavy English influence and the effects of globalization. The language focuses on Cebuano, the second most spoken language mainly spoken in Central Visayas. Existing Cebuano parsers have been made through Recursive Descent Parsing (RDP) for a Cebuano Parse Tree in relaying grammar through a syntactic structure. However, there were identified limitations through low-level sentences, lacking grammar codes, low metric score, and stemmer evaluation errors. To address this gap, the researchers proposed a Snowball-based stemmer approach along with the Cocke-Kasami-Younger algorithm and additional grammar codes for better syntax tree optimization. Due to the lack of a dataset for Cebuano sentences, the Cebuano Corpus is extensively created by the researchers via web scraping and data augmentation techniques such as language dialect filtration and sentence range variation. The dataset comprises at least 500 collected phrases to be annotated by the linguist experts for the Gold Standard Parse Treebank (GSPT). The GSPT will serve as the parser's benchmark through syntax constituents, known as the PARSEVAL metric system. The improved parser attained an accuracy of 82.14%, suggesting a high rate of correct Cebuano constituents from a Cebuano sentence. The findings suggest that the developed system is more effective than its predecessor. By leveraging advanced Snowball-based algorithms and updated Grammar Codes, the study demonstrates the potential of parsers in preserving cultural language.
The article explores the key aspects of language functioning in modern media texts, focusing on its specific features, stylistic characteristics, and impact on the audience. It also examines the role of language as a tool for communication, manipulation, and shaping public opinion, while addressing the transformation of language norms in the context of digitalization and the emergence of new media formats. In the modern information society, media texts are recognized as an essential part of public life. They play a significant role in shaping public opinion, influencing political preferences, and promoting cultural values. The language of media texts acts as an indicator of societal processes, reflecting changes in methods of information dissemination, linguistic practices, and the growing demand for interactivity among audiences. The study emphasizes the language of media as fulfilling not only an informative function but also aesthetic, persuasive, emotional, and manipulative ones, among others. It highlights that the language of media texts mirrors complex social and technological processes within society, opens up new opportunities for communication, and demands responsibility and attention to societal impacts. Analyzing the current issues in the language of modern media texts provides a deeper understanding of their influence on society and potential pathways for future development. The article also underlines that media texts employ specific linguistic tools to create emotional impact and facilitate manipulation, influencing not only the behavior but also the beliefs of their audience. Furthermore, the language of media texts contributes to the transformation of linguistic norms, introducing new terms, slang, abbreviations, and hybrid constructions that reflect the dynamic evolution of society. Key features of the language in modern media texts include hybridization, the use of colloquial elements, manipulation of consciousness, interactivity, and multifunctionality. Stylistic features such as conciseness, brevity, emotionality, and expressiveness are identified as essential components of modern communication. Additionally, the use of various stylistic devices further enhances the effectiveness and appeal of media texts.
Modern natural language processing is fundamentally reliant on word embeddings, which represent words as dense vectors in a high-dimensional space. While the utility of the vector-space structure is well-established, particularly through linear algebraic analogies for semantic relations, the richer geometric properties of this space remain largely unexplored. This paper investigates the intrinsic geometry of word embedding spaces, positing that local manifold curvature serves as a quantitative proxy for semantic ambiguity. We hypothesize that polysemous words, which act as conceptual bridges between disparate semantic contexts, induce a higher degree of local curvature in the embedding manifold. To test this, we model the local neighborhood of each word vector as a discrete graph and compute its Ollivier-Ricci curvature. This geometric measure is then correlated with a standard linguistic measure of semantic ambiguity, namely the number of distinct synsets for each word as cataloged in the WordNet lexical database. Our experiments, conducted on widely-used pre-trained GloVe embeddings, reveal a statistically significant positive correlation between the calculated local curvature and the degree of polysemy. Words with high ambiguity, such as 'bank' or 'lead', consistently exhibit greater curvature than semantically stable words like 'microscope' or 'hydrogen'. These findings suggest that the geometric structure of embedding spaces encodes sophisticated semantic properties beyond simple vector proximity. The quantification of semantic ambiguity through curvature offers a novel, model-agnostic tool for analyzing embedding quality and has potential applications in improving word sense disambiguation systems and developing more robust language models.
This study addresses the increasing prevalence of emotional and behavioral problems among adolescents by designing and validating a game-based intervention model grounded in the Process Model of Emotion Regulation. We propose a novel integration of biofeedback therapy with Rational Emotive Behavior Therapy to establish a three-stage intervention framework: “emotional exposure - biofeedback - strategy reinforcement”. The research introduces Balance Masters, a serious game featuring a platform balancing mechanic as its core gameplay, connected to a Bluetooth heart rate sensor to implement a real-time biofeedback system. Through a self-controlled experiment with adolescents in elementary schools, results demonstrated significant improvements in cognitive reappraisal scores and significant reductions in expressive suppression scores post-intervention. Heart rate measurements showed a significant decreasing trend throughout gameplay sessions, with particularly notable differences between late and early phases. Game experience evaluations indicated high levels of immersion and positive affect ratings with strong usability metrics. This research contributes an accessible, highcompliance innovative solution for addressing emotion regulation challenges in adolescents, with implications for digital mental health interventions.
Ce travail de recherche se penche sur l’analyse contrastive de l’annotation syntaxique de certaines constructions complexes en arabe et en français, réalisée selon le schéma Universal Dependencies (UD) et de sa variante Surface-Syntactic Universal Dependencies (SUD). L’analyse portera sur trois catégories de constructions: les constructions relatives, les constructions à verbe support et les constructions copulatives et avec un auxiliaire. La recherche a un double but: d’une part, elle vise à explorer comment les schémas UD et SUD représentent ces constructions syntaxiques en arabe et en français. D’autre part, il s’agira d’évaluer la pertinence et la capacité de ces schémas à rendre compte des particularités propres à chaque langue. Les résultats révèlent que les annotations syntaxiques, bien que cohérentes pour les cas standards, divergent dans les constructions non canoniques, une incohérence constatée tant entre les corpus arborés arabe et français qu’au sein de chacun d’eux. Des solutions s’imposent: harmoniser les treebanks arabes et français existants par la standardisation et la correction, et optimiser les guides d’annotation pour les langues concernées.
Abstract Facial expressions provide rapid and informative cues about others’ emotional and mental states, playing a critical role in social interactions. However, whether distinct emotional expressions reflect discrete neural processes or arise from varying combinations of underlying affective dimensions such as arousal and valence remains a subject of investigation. Crucially, these accounts need not be mutually exclusive: different brain regions may encode emotional expressions along both categorical and dimensional axes to varying degrees. To test this hypothesis, we probed different brain systems involved in emotion recognition - the ventral attentional network and the cortical limbic system centered on the ventromedial prefrontal cortex (vmPFC) - to investigate the extent to which these networks and their subregions encode emotional facial expressions in terms of (1) perceived arousal, (2) arousal+valence, or (3) six discrete emotion categories (anger, disgust, fear, happiness, pain, sadness). To this aim, we modelled the fMRI signal from perceiving movies of facial emotional expressions with ratings for either arousal, arousal+valence or emotion category using representational similarity analysis (RSA) - a method that aims at assessing which ratings model better represents how the perception of facial expressions is reflected by the fMRI parameter estimates across all the voxels within a brain region. This analysis showed that regions in the vmPFC network, including the subgenual cingulate and the medial OFC, are sensitive to ratings for distinct emotion category and for arousal+valence - significantly more so for the former - while they fail to show sensitivity for arousal ratings alone. In the ventral attentional network, the mid-posterior insula showed a similar profile, while the most posterior and ventral insular show evidence of encoding arousal+valence, but not for emotion category or arousal alone. Our findings support the idea that the vmPFC and the mid-posterior insula play a role in encoding emotion-specific and valence representations beyond general arousal processing. These results contribute to understanding how integrative brain regions support emotional empathy and social cognition.
This article explores the requirements for corpus compilation within the GiesKaNe project (University of Giessen and Kassel, Syntactic Basic Structures of New High German). The project is defined by three central characteristics: it is a reference corpus, a historical corpus, and a syntactically deeply annotated treebank. As a historical corpus, GiesKaNe aims to establish connections with both historical and contemporary corpora, ensuring its relevance across temporal and linguistic contexts. The compilation process strikes the balance between innovation and adherence to standards, addressing both internal project goals and the broader interests of the research community. The methodological complexity of such a project is managed through a complementary interplay of human expertise and machine-assisted processes. The article discusses foundational topics such as tokenization, normalization, sentence definition, tagging, parsing, and inter-annotator agreement, alongside advanced considerations. These include comparisons between grammatical models, annotation schemas, and established de facto annotation standards as well as the integration of human and machine collaboration. Notably, a novel method for machine-assisted classification of texts along the continuum of conceptual orality and literacy is proposed, offering new perspectives on text selection. Furthermore, the article introduces an approach to deriving de facto standard annotations from existing ones, mediating between standardization and innovation. In the course of describing the workflow the article demonstrates that even ambitious projects like GiesKaNe can be effectively implemented using existing research infrastructure, requiring no specialized annotation tools. Instead, it is shown that the workflow can be based on the strategic use of a simple spreadsheet and integrates the capabilities of the existing infrastructure.
This research paper proposes a conceptual framework and optimization algorithm for pruning techniques in deep learning models, its focus is on key challenges such as model size, computational efficiency, inference speed and sustainable technology development. The aim of the framework is to transition from large neural networks to sparse, efficient models, indicating the benefits of pruning in improving model scalability and applicability of the pruned models. The proposed framework focuses on reducing the model size, optimizing training schedules and facilitating efficient deployment in real-world devices. The development of the framework involves four stages: Reviewing critical research concepts, identifying relationships between concepts and designing the pruning framework. Furthermore, this study also introduces a new multi-objective optimization algorithm that formalizes the trade-offs between accuracy, computational cost, inference time and energy consumption in the pruning process. Our experiments demonstrate the method's effectiveness in achieving notable model compression while preserving competitive performance on a sentiment analysis and linguistic acceptability tasks using Stanford Sentiment Treebank (SST-2) and Corpus of Linguistic Acceptability (CoLA) datasets. The results of our experiments show the BERT Base model being pruned to 25 million parameters gaining an accuracy of 96.3% on SST-2 dataset and F1-score of 95.2%. Furthermore, the pruned model demonstrated F1 score of 82.3 and 56% of Matthews correlation coefficient in CoLA dataset respectively. This framework, along with the algorithm, serves as a reference for researchers and practitioners, who can select a suitable approach based on the specific application requirements for pruning deep learning models.
This study examines the intersection of language, identity, and politics in Nigerian online discussion through a sociolinguistic analysis of Nairaland, Nigeria's largest internet forum. Drawing on Communication Accommodation Theory (CAT) and sociolinguistic theory, the research investigates how participants strategically use code-switching, Nigerian Pidgin, and local languages to negotiate identity, voice dissent, and establish community in politicised debates. Three selected threads, "Igbo Founded Ile-Ife," "Okrika Women Protest Unlawful Suspension of Fubara," and "Rivers Emergency Rule", serve as case studies for the analysis of how language works as a vehicle to convey ethnic pride, political grievance, and gendered opinion in online environments.Participants are observed to employ convergence, divergence, and maintenance strategies to show group membership, claim differences, or refuse linguistic norms. Nairaland's language options are not arbitrary but serve affective, rhetorical, and ideological functions, permitting speakers to negotiate group membership or sociopolitical grievances. The study reveals how Pidgin English and blended varieties of language are employed to construct a collective sense of identity and solidarity, particularly in protest or satirical contexts. It also discloses the gendered politics of exclusion whereby women's political voices are diminished through sexist and class-based labelling, an outcome of more profound societal biases replicated within the virtual spaces.The study contributes to the knowledge of digital multilingualism and bottom-up creativity in African online forums, highlighting the importance of ethical digital literacy and the moderation of inclusive discourses. By examining language use on Nairaland, the book enlightens readers on how Nigerians imaginatively appropriate linguistic material to shape power relations, assert their identities, and construct solidarity in the digital sphere. Relevant to scholars in sociolinguistics, digital humanities, and African studies, the book presents a case study for examining political discourse and identity creation in multilingually complex cyberspaces.
State-space models (SSMs) have recently emerged as efficient alternatives to computationally intensive architectures such as Transformers for sequence modeling. However, their training typically relies on static loss functions, which may be suboptimal at different stages of learning. In this work, we introduce a hybrid model that integrates the Hyena architecture with a Dynamic Loss Network (DLN) under a Learning-to-Teach (L2T) paradigm, referred to as L2T-DLN. In this framework, the Hyena model serves as a student whose loss function is adapted online, while a teacher model, equipped with a memory of the student's past performance, guides the DLN to dynamically trade off the primary cross-entropy objective and a regularization term. We evaluate the proposed L2T-Hyena model on the Penn Treebank (PTB) and WikiText-103 language modeling benchmarks and compare it against both a vanilla Hyena SSM and a Transformer baseline. On PTB, our model achieves a validation perplexity of 102.6, representing a substantial improvement over the 110.5 obtained by the vanilla Hyena trained with a static loss function and 121.28 achieved by the Transformer baseline. Similar gains are observed on WikiText-103, where L2T-Hyena reaches a validation perplexity of 68.3, outperforming vanilla Hyena (73.7) and Transformer (89.8). These results indicate that coupling SSMs with adaptive loss functions can significantly enhance both the quality and efficiency of deep learning models for sequential data and hold strong promise for applications in natural language processing, time-series analysis, and biological signal processing.
This article utilizes contextual semantic description technology, the first step in linguistic database preparation for building a corpus of religious texts. The book "Sahih ul-Bukhari" was chosen as the study's subject for this. The scientific contribution of the article is that it explains the processes of manual and automatic database preparation. The advantages of the contextual semantic description technique are explained, namely for religious and classical texts. This study examines the semantic fields and categorical classifications of 428 lexical units extracted from the referenced corpus, utilizing Mukhammad Sadiq Mukhammad Yusuf’s "Sakhih ul-Bukhari" (1-juz, 2019) as its primary source, and provides a comprehensive account of their semantic annotation. Special emphasis is placed on the high-frequency terms "Book" and "Hadith," which are analyzed through thematic grouping, semantic tagging, and examination of keyword-based co-occurrences. This study identified 18 distinct thematic categories. The research utilizes methodological approaches such as keyword and collocational analysis, thematic modeling, and semantic tagging to examine the corpus. The methods used in the study include thematic modeling, semantic tagging, and contextual analysis of lexical units. The discussion of semantic challenges and descriptive methodologies presented here contributes significantly to the analytical paradigms within corpus linguistics.
Introduction Neurodegenerative diseases like progressive supranuclear palsy (PSP) present challenges concerning their diagnosis. Neuroimaging using magnetic resonance (MRI) may add diagnostic value. However, modern techniques such as volumetric assessment using Voxel Based Morphometry (VBM), although proven to be more accurate and superior methods compared to MRI, have not gained popularity among scientists in the investigation of neurological disorders, due to their higher cost and time‐consuming applications. Conventional brain MRI methods may present a quick, practical, and easy to use imaging rating tool for the differential diagnosis of PSP. The purpose of this study is to evaluate a string of existing visual MRI rating scales and signs with regards to their impact for the diagnosis of PSP. Materials and Methods The population study consisted of 30 patients suffering from PSP and 72 healthy controls. Each study participant underwent a brain MRI, which was subsequently examined by two independent researchers, in a double-blinded fashion. Fifteen visual rating scales and signs were evaluated including pontine atrophy, cerebellar atrophy, midbrain atrophy, aqueduct of Sylvius enlargement, cerebellar peduncles hyperintensities, οκι8θ of the fourth ventricle (100% sensitivity ‐ 71% specificity) and left temporal lobe atrophy (97% sensitivity ‐ 78% specificity). Conclusions Enlargement of Sylvius aqueduct, enlargement of the fourth ventricle and atrophy of both temporal lobes together with presence of morning glory and hummingbird signs can be easily and quickly distinguished and identified by an experienced radiologist without involving any complex analysis making them a useful tool for the PSP diagnosis. MRI visual scale measurements could be added to the diagnostic criteria of PSP and may serve as an alternative to highly technical and more sophisticated quantification methods.
The intricate interplay between visual perception and emotion determines how waking experience influences mentation through a 'day residue' at once conspicuous yet hard to predict. Here we set out to map the neural sources associated with the visuo-affective processing of the 'day residue' during hypnagogic sleep. To this end, we assessed 28 healthy participants on a combined nap protocol with serial awakenings, pre-sleep stimulation with affective visual images, yoked measures of the semantic similarity between image and imagery reports, affect ratings, estimation of 64-channel EEG sources, and functional connectivity analysis. Overall, low-frequency EEG power was associated with weaker residues, and high-frequency EEG power was associated with stronger residues. The source networks most significantly correlated with imagetic and affective residues were markedly different across wake-sleep states, partially overlapping with the default mode network during N1 for up to 50% and 61%, respectively. The results allowed us to identify neural correlates of the visuo-affective processing of the day's residue, showing that the hypnagogic processing of the waking experience involves complex, dynamic and sequential bi-hemispheric interactions among multiple cortical, subcortical, and cerebellar structures with visual, limbic, optokinetic, and cognitive functions.
Do images in corporate disclosures convey genuine information or merely embellishment?&nbsp; We advance the visual informativeness literature by introducing a framework delineating how images reinforce text and what properties of such reinforcement matter. Our methods apply wherever images and text coexist. We classify reinforcement as substantive or hype — operational or merely evocative. Testing in an image-dense setting (a large sample of U.S. sustainability reports), we find that reinforcing image pages predict expert ratings more strongly than non-reinforcing ones, with the effects amplified by proximity, potency, and sentence importance. Sophisticated raters respond more strongly to substantive than hype reinforcement. PCA-based composite indices provide robust validation, and the vehicle for causal identification within a quasi-experimental setting: TruValue ratings — machine-driven and blind to company-issued images, unlike Refinitiv ratings — reveal no association with images, suggesting that images affect ratings through a human processing channel, not through a machine algorithm that lacks the human-image interaction. <br> <br> This paper was previously circulated under the title: <b>"Going the Distance:" Images in Corporate ESG Reports</b>
BACKGROUND: Lamotrigine has been shown to be effective in the long-term treatment and relapse prevention of depression in bipolar disorder. However, the neuropsychological mechanisms underlying these effects are unclear. We investigated the effects of lamotrigine on a battery of emotional processing tasks in healthy volunteers, previously shown to be sensitive to antidepressant drug action in similar experimental designs. METHODS: Healthy volunteers (n = 36) were randomized in a double-blind design to receive a single dose of placebo or 300 mg lamotrigine. Mood and subjective effects were monitored throughout the study period, and emotional processing was assessed using the Oxford Emotional Test Battery (ETB) 3 hours post-administration. RESULTS: Participants receiving lamotrigine showed increased accurate recall of positive versus negative self-descriptors, compared to those in the placebo group. There were no other significant effects on emotional processing in the ETB, and lamotrigine did not affect ratings of mood or subjective experience. CONCLUSIONS: Lamotrigine did not induce widespread changes in emotional processing. However, there was increased positive bias in emotional memory, similar to the effects of antidepressants reported in previous studies. Further work is needed to assess whether similar effects are seen in the clinical treatment of patients with bipolar disorder and the extent to which this is associated with its clinical action in relapse prevention.
Without social communication, people's daily lives and any of their activities are entirely unimaginable. A specific form of social communication is verbal interaction, despite modern psychological data indicating that 65% of information in communication is conveyed through nonverbal means. Every specific act of verbal communication carries a pragmatic orientation toward the partner involved in the interaction. Verbal communication inherently involves the execution of speech acts, which also includes the use of speech etiquette. In reality, it is a valuable component of any human behavior, significantly influencing the effectiveness of its completion. People need shared knowledge of etiquette, including speech etiquette norms, in the communication process to successfully establish effective communication. Etiquette is primarily a social phenomenon, but it is also a cultural, historical, and psychological one. It inevitably has a linguistic aspect—more specifically, linguistic etiquette, which refers to the set of linguistic norms characteristic of communication in various situations. In specialized literature, there is an opinion that insufficient attention is given to the psychological nature of speech as a specific type of activity. However, it has long been a subject of special study in psycholinguistics (e.g., Austin's theory of speech acts, 1955). According to D. Uznadze’s model of set structure, known as the static model, speech activity is considered an independent behavior, defined by its own autonomous set. It does not matter whether it arises from a necessity within another ongoing behavior or is directly driven by the need for verbal communication. In contrast, the dynamic model of speech activity and, consequently, speech etiquette, allows us to view it as an activity shaped by the need for linguistic communication, i.e., linguistic set. This activity constitutes a significant component of behavior, adapting dynamically to the primary attitude of that behavior. In this model, the tendency for permanent modification is embedded from the outset, eliminating the need for the destruction or reconstruction of sets. It considers both practical and theoretical activities directed toward a common goal as actions integrated into the same behavior, operating based on its fundamental set.
Speech Emotion Recognition (SER) plays a pivotal role in understanding human communication, enabling emotionally intelligent systems, and serving as a fundamental component in the development of Artificial General Intelligence (AGI). However, deploying SER in real-world, spontaneous, and low-resource scenarios remains a significant challenge due to the complexity of emotional expression and the limitations of current speech and language technologies. This thesis investigates the integration of Automatic Speech Recognition (ASR) into SER, with the goal of enhancing the robustness, scalability, and practical applicability of emotion recognition from spoken language. As a starting point, we explore the interplay between ASR and SER by conducting an in-depth analysis of speech foundation models on emotionally expressive speech, from both acoustic and linguistic perspectives. Our findings uncover inherent limitations of these models: while they achieve strong performance in ASR, they often fail to capture paralinguistic cues essential for emotion recognition, and exhibit emotion-dependent biases. Additionally, we examine how linguistic properties, such as part-of-speech distributions, affective word ratings, and utterance length, interact with ASR performance across different emotion categories. Extensive experiments reveal that ASR errors are not random but follow systematic patterns associated with specific word types and emotional expressions across diverse speaking conditions. To overcome paralinguistic limitations while utilizing the hierarchical encoding capabilities of speech foundation models, we propose a joint training framework that integrates ASR hidden representations and lexical outputs in a hierarchical manner for SER. To mitigate transcription quality issues, we introduce two complementary ASR error correction strategies that require only a small amount of emotional speech data: a Large Language Model (LLM)-based method that refines N-best hypotheses using emotion-specific prompts, and a Sequence-to- Sequence (S2S) model that leverages discrete acoustic units to correct 1-best hypothesis. Both approaches substantially improve the emotional accuracy of ASR transcriptions with low-resource speech data and enhance downstream SER performance. Beyond transcription correction, this thesis introduces robust acoustic-lexical fusion frameworks to handle emotion mismatch across modalities and reduce ASR error propagation. We conduct a comprehensive investigation into cross-modal incongruity and propose incongruityaware fusion strategies that dynamically adapt to modality conflicts, extending solutions beyond SER to include sarcasm and humor detection. Building on this, we develop an ASRii aware, modality-gated fusion mechanism that integrates ASR error correction with dynamic modality selection in a sequential manner. These fusion strategies achieve strong performance on both controlled and real-world datasets, even under partially erroneous conditions (i.e. ASR transcriptions). Finally, to reduce reliance on costly labeled emotion data, we present a novel semi-supervised learning framework based on multi-view pseudo-labeling. By leveraging both acoustic similarity metrics and LLM-based confidence estimation, this approach selects high-quality unlabeled samples to augment training. The proposed framework not only outperforms traditional pseudo-labeling methods for SER but also demonstrates generalizability to speechbased Alzheimer’s dementia detection.
OBJECTIVES: If task-irrelevant sounds are present when someone is actively listening to speech, the irrelevant sounds can cause distraction, reducing word recognition performance and increasing listening effort. In some previous investigations into auditory distraction, the task-irrelevant stimuli were non-speech sounds (e.g., laughter, animal sounds, music), which are known to elicit a variety of emotional responses. Variations in the emotional response to a task-irrelevant sound could influence the distraction effect. The goal of this study was to examine the relationship between the arousal (exciting versus calming) or valence (positive versus negative) of task-irrelevant auditory stimuli and auditory distraction. Using non-speech sounds that have been used previously in a distraction task, we sought to determine whether stimulus characteristics of arousal or valence affected word recognition or verbal response times (which serve as a measure of behavioral listening effort). We anticipated that the perceived arousal and valence of task-irrelevant stimuli would be related to distraction from target stimuli. DESIGN: In an online listening task, 19 young adult listeners rated the valence and arousal of non-speech sounds, which previously served as task-irrelevant stimuli in studies of auditory distraction. Word recognition and verbal response time data from these previous studies were reanalyzed using the present data to evaluate the effect of valence or arousal stimulus category on the distraction effect in quiet and in noise. In addition, correlation analyses were conducted between ratings of valence, ratings of arousal, word recognition performance, and verbal response times. RESULTS: The presence of task-irrelevant stimuli affected word recognition performance. This effect was observed generally in quiet and for stimuli rated as exciting (in noise) or calming (in quiet). The presence of task-irrelevant stimuli also affected reaction times. Background noise increased verbal response times by approximately 35 msec. In addition, all task-irrelevant stimuli, regardless of valence or arousal category, increased verbal response times by more than 200 msec relative to the condition with no task-irrelevant stimuli. Valenced stimuli caused the largest distraction effect on response times; there was no difference in the distraction effect on verbal response times based on the stimulus arousal category. Correlation analyses between valence ratings and dependent variables (word recognition and reaction time) revealed that, in quiet, there was a weak, but statistically significant, relationship between valence ratings (absolute deviation from neutral) and word recognition scores; the more valenced a stimulus, the more distracting it was in terms of word recognition performance. This significant relationship between valence and word recognition was not evident when participants completed the speech task in noise. There was no relationship between stimulus ratings (arousal or valence) and reaction time in quiet or in noise. CONCLUSIONS: Valenced (positive and negative) or exciting task-irrelevant non-speech sounds can negatively affect word recognition and increase listening effort. Future study should consider the emotional content of task-irrelevant stimuli when evaluating potential distraction effects.
Introduction: Receptive Music Therapy allows individuals with sub-clinical anxiety levels to self-medicate when and where they choose, but the effectiveness of self-administered 'music medicine' to enhance psychological well-being is still being investigated. The current study reports on a song ('Bagels') designed to alleviate mild anxiety in adolescents and young adults. Methods: A laboratory study was conducted to examine the effect of Bagels on brain states, and upon both subjective and objective measures of state anxiety. Measures of skin conductance and heart rate (HR), and 64-channel Electroencephalography (EEG) were obtained from 30 young adults as they listened to six songs contrasting affective properties. Subjective measures included ratings of a song's pleasantness, arousal, dominance, and likability, as well as estimates of state anxiety obtained immediately after listening to them. Results: Preliminary analyses revealed that the six songs differed significantly in terms of affective ratings, with Bagels rated as more pleasant and less arousing, and having lower state anxiety ratings at its terminus. EEG alpha connectivity was also lowest for the song Bagels, particularly in the brain's frontal lobes. Similarly, Bagels was associated with lower physiological arousal, reflecting less arousal and greater calmness. Discussion: Combined, the analysis suggest that Bagels has the potential to be an effective digital anxiolytic. Discussion around the promise of music medicine and aspects of its management are presented, along with avenues of further inquiry.
This paper introduces a Bias-aware Text Mining: System, a novel approach to address challenges in Bias-Aware Text Mining: A Novel Framework for Fair and Transparent Language Models. Our BTMS framework leverages advanced algorithms to improve the performance metrics by approximately <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{4 1 \%}$</tex> compared to existing methods. Experiments conducted on standard datasets demonstrated the effectiveness of the proposed approach, particularly in terms of accuracy. The proposed system integrates multiple computational techniques, including group theory, combinatorial optimization, and knowledge distillation, to create a robust solution that outperforms current state-of-the-art methods. Specifically, a comprehensive evaluation using the PENN Treebank and WMT-14 demonstrates that BTMS achieves superior performance across multiple evaluation criteria. Our establishment addresses key limitations in existing techniques by incorporating an adaptive learning rate and cross-domain adaptation, which enables more effective handling of complex data patterns. The experimental results confirm that our methodology reduces the computational complexity while maintaining high accuracy, making it suitable for real-world applications with resource constraints. We also conducted ablation studies to analyze the contribution of each component to the overall performance (a key finding), revealing that the embedding module is particularly critical for achieving optimal results. In addition, we performed a sensitivity analysis to assess the robustness of the BTMS under varying conditions, confirming its stability across different operational scenarios. Importantly, the theoretical analysis provides formal (notably) guarantees of the convergence properties and computational efficiency of our algorithm. Finally, we discuss the potential applications of our approach in related domains and outline directions for future research to enhance the capabilities of the proposed system further.
Our socio-emotional development and well-being critically depends on interpersonal tactile interactions, which are sensed by the skin through C-tactile (CT) afferents that respond to gentle, slow touch at typical skin temperatures. In the present study, we investigated whether impersonal touch would be able to provide similar pleasantness compared to interpersonal touch within a body-ownership illusion paradigm. To provide impersonal touch at similar parameters, we used a thermal probe kept at [Formula: see text]C (typical skin-to-skin temperature) compared to a flat hand as interpersonal touch. Both forms of touch were performed at CT-compatible speeds of 3cm/s by a male trained experimenter within a classic rubber hand illusion (RHI) paradigm in two counter-balanced within-participant conditions. A sample of N=45 healthy participants was tested and pleasantness ratings, touch deprivation, and the Need-For-Touch-Scale (NFT) were gathered. Overall, the illusion was similar in both touch conditions and, importantly, we found no statistically significant difference in pleasantness between interpersonal and impersonal touch. Interestingly, neither NFT scores, nor touch deprivation measures correlated with individual differences in the RHI and affective ratings. Our results suggest that impersonal touch with CT-optimal components provide a pleasantness and subjective illusion experience compared to interpersonal touch under the RHI paradigm.
• One-word text messages with periods are seen as more negative than texts without periods. • We replicated these negative effects of message-final periods in one-word messages. • The negative effects of periods extended to 3–4-word, but not 6–8-word messages. • Increasing message length led to more positive message ratings. • Punctuation in text messages can convey rhetorical and grammatical information. Including a period at the end of a one-word text message (“yes.”) can elicit perceptions of insincerity and negativity. We assessed whether these negative perceptions would hold in longer messages. Australian undergraduates ( N = 200) read 30 fictitious message exchanges; positive, neutral, or negative in valence. Exchanges ended in a short (one word), medium (three to four words), or long (six to eight words) message; half with a final period, half without. Participants rated how they thought the friend felt about their message, on a 7-point scale. Across valences, messages with a period were rated more negatively than those without, as were both short and medium messages. In long messages, the period’s presence or absence did not significantly affect ratings. The results imply that this punctuation mark can convey emotional or grammatical information, depending on message length. The findings have implications for the way people compose and interpret digital messages.
BACKGROUND: Attractive individuals receive higher ratings on independent social attributes including intelligence, success, and likability. The authors sought to determine whether knowledge that a person in a photograph had facial plastic surgery affects perceptions of attractiveness or social attributes. METHODS: US adults were recruited to complete an online survey (Qualtrics respondent distribution system). Demographic information, a cosmetic surgery questionnaire, and a body dysmorphic disorder screening questionnaire were collected. Participants viewed photographs of young, White, male- and female-presenting individuals with above-average attractiveness (selected by Delphi method). Half of the participants were told that the individuals had facial plastic surgery. All participants rated the people presented in photographs in terms of attractiveness and social attributes. Descriptive statistics were used to analyze the data. RESULTS: A total of 1000 participants (500 female; mean age, 49 years [SD 16.9]) were enrolled. There were no differences between participants in the primed ( n = 500; 250 female) or unprimed groups based on demographics, bias toward cosmetic procedures, or body dysmorphia screen. Ratings of the attractiveness and social attributes of people presented in photographs did not differ between the primed and unprimed groups (all P > 0.13). This held true even for those with negative bias toward cosmetic surgery. CONCLUSION: Knowledge that someone had facial plastic surgery does not affect ratings of attractiveness or social attributes.