From linguistic analyses to large language models: A scoping review of methods used to investigate language features in depression research.

Infants develop the ability to segment words from continuous speech from the age of 6 months onwards. However, in French- and English-learning infants, this ability initially appears to be restricted to consonant-initial words. Vowel-initial word segmentation only emerges at around 13.5 months of age, or at 11 months when additional segmentation cues are provided. Previous research has attributed this asymmetry to phonological and acoustic differences between word-initial consonants and vowels. The present study examined whether distributional properties such as the likelihood of encountering vowel-initial words also contribute to this developmental pattern. For this, 114 infants learning German, a language with a higher proportion of vowel-initial words than English and French, were tested. Using the head-turn preference procedure, 11-month-olds (Experiment 1) and 8-month-olds (Experiment 2) were familiarized with isolated words beginning with either consonants or vowels. Subsequently, their preference for listening to text passages containing these words was assessed. While the 8-month-olds showed no evidence of segmentation for either word type, the 11-month-olds successfully segmented both consonant- and vowel-initial words, showing no advantage for either of the word types. The results support an input-based account: Experience with a language richer in vowel-initial words may facilitate earlier segmentation of vowel-initial forms. Overall, the findings highlight the influence of language-specific distributional properties on early segmentation biases and underscore the need for research with infants learning structurally diverse languages.

The primary challenge in deploying Large Language Model (LLM) is ensuring its harmlessness. Red team can identify vulnerabilities by attacking LLM to attain safety. However, current efforts heavily rely on single-round prompt designs and unilateral red team optimizations against fixed blue teams. These static approaches lead to significant reductions in generation diversity, known as the mode collapse, which makes it difficult to discover the potential risks in the increasingly complex human-LLM interactions. Here we introduce dynamic Red Team Game (RTG) to comprehensively analyze the multi-round offensive and defensive interactions between red team and blue team. Furthermore, we develop a Gamified Red Team Solver (GRTS) with diversity measures to mitigate mode collapse and theoretically guarantee the convergence of approximate Nash equilibrium which results in better strategies for both teams. Empirical results demonstrate that GRTS explore diverse and implicit attacks to adaptively exploit various LLMs, surpassing the constraints of specific modes. Insightfully, the geometrical structure we unveil of the red team task aligns with the spinning top hypothesis, confirming the necessity of constructing a diverse LLM population as a promising proxy for heterogeneous human expert red-teamers. This paves the way for scalable toxicity detection and safe alignment for LLMs.

There is an emerging understanding of diverse language preferences in autism among stakeholders, including that for identity-first language (IFL) versus person-first language (PFL). We aimed to understand preferences regarding language use in racially and culturally diverse Singapore. A cross-sectional study was completed by autistic adults, caregivers of autistic individuals, and professionals working in the autism field. The questionnaire comprised commonly used terms in autism, which respondents rated from 1 (uncomfortable/I do not like this term) to 7 (very comfortable/this would be my preference). The sample comprised 320 individuals (82.8% females; Chinese 75.9%, Malay 9.4%, Indian 8.8%). Respondents comprised 38 (11.9%) autistic individuals, 86 (26.9%) caregivers/relatives, and 196 (61.3%) professionals. Differences in language preferences were observed between respondent types. Autistic individuals preferred terms like 'autistic' (M ± SD = 5.00 ± 2.25) unlike caregivers/relatives (3.66 ± 2.41) or professionals (3.30 ± 1.94, p < .001). PFL terms like 'person with autism' and 'person with ASD' were preferred by caregivers/relatives (4.47 ± 2.19; 4.33 ± 2.25) and especially professionals (5.32 ± 1.58; 4.89 ± 1.84) but disliked by autistic individuals (3.11 ± 2.09; 2.39 ± 1.99; p < .001). Different language preferences were observed between those with lived experience versus professionals. Sensitivity to the preferences of stakeholders fosters better inclusion and acceptance of neurodiversity.Lay AbstractLanguage shapes how autism is perceived and understood by society. Research on language preferences in the autism community has been mostly from Western sources. The primary consideration has been in identifying the preference between identity-first language (IFL), such as 'autistic person', and person-first language (PFL), such as 'person with autism'. These studies suggest that professionals and caregivers prefer PFL while autistic individuals favor IFL. Little is known about these preferences in a culturally diverse setting like Singapore. This study aimed to explore the language preferences among the stakeholders in the autism community in Singapore. In this study, 320 participants were surveyed anonymously online; participants were asked to rate their comfort level with certain autism-related terms. The survey findings revealed significant differences in preferred language-autistic individuals expressed a strong preference for IFL and for terms that conveyed autism as a part of their identity and did not prefer clinical terms such as 'symptoms of autism' or PFL terms such as 'person with ASD'. In contrast, caregivers and professionals tended to prefer PFL terms and medicalized language. It is likely that societal factors, including Singapore's relatively conservative culture and societal views on autism, influenced these preferences. Awareness of the variations in language preferences among stakeholders in the autism community will help to foster inclusivity, acceptance, and sensitivity toward autistic individuals and their families.

Informed consent is a critical component of patient-centred care, yet variability in its delivery can undermine patient understanding, satisfaction and autonomy. In practice, however, the process is often hindered by high clinical workloads, time pressures and the absence of a standardised approach, which can lead to incomplete explanations, variability in practice and reduced patient comprehension. This project aimed to improve patient understanding and satisfaction with bedside procedures by introducing a standardised informed consent process.This study was conducted at Hamad General Hospital’s Acute Medical Assessment Unit to enhance the informed consent process for routine medical procedures, including thoracocentesis, paracentesis, lumbar puncture and blood transfusions. Prefilled consent forms were developed for each bedside procedure, detailing both common and uncommon complications with estimated prevalence. Multilingual patient information templates and pictorial aids were produced to address the needs of patients from diverse language and literacy backgrounds. The intervention was implemented over 10 sequential Plan–Do–Study–Act cycles, each lasting 4 weeks, targeting barriers such as workflow integration, accessibility of forms and staff engagement.Patient satisfaction with the informed consent process increased from 64% to 94% over the course of the project. Patients reported improved understanding of procedure risks, benefits and alternatives and nursing staff described greater confidence and proactive participation in the process.This outcome supports a structured, standardised consent process supported by multilingual written materials and visual aids, which can substantially improve patient understanding and satisfaction for bedside procedures. Embedding these tools into routine workflows, supported by multidisciplinary engagement, can deliver more equitable and consistent patient-centred care.

Indonesia is a multicultural nation-state, characterized by diverse ethnic groups, languages, religions, cultures, traditions, and regional identities. Therefore, developing instruments to assess critical thinking skills based on multicultural values (CTSMV) for history teachers to be implemented in the learning process is highly important. This study aims to: (1) analyze the development of the factual model for constructing the instrument; (2) evaluate the design of the development model for constructing CTSMV instruments; and (3) evaluate the feasibility of the developed CTSMV instrument construction model for history teachers. The research design employed Research and Development (R&amp;D) using the Borg and Gall model consisting of ten steps, which were simplified into three stages: (1) preliminary study; (2) design development and evaluation stage; and (3) product effectiveness testing stage. The research subjects consisted of 50 junior high school and senior/vocational high school history teachers. Data were collected through observation, interviews, and questionnaires, while data analysis applied both quantitative and qualitative approaches. The findings reveal that: (1) the training management model for instrument construction implemented so far has not adequately accommodated teachers’ needs and aspirations; (2) the design of the CTSMV instrument construction training management model was developed based on factual conditions, literature review, previous research findings, and training needs analysis; and (3) the evaluation results indicate that the CTSMV instrument construction training management model is feasible to be used and implemented by history teachers in the learning process. The model design represents a developed product validated by experts and practitioners, and it has been tested through individual trials involving five participants, group trials involving ten participants, and limited trials involving thirty-five participants, all of which produced good evaluation criteria. Therefore, the resulting product development model is appropriately named the “Training Management Model for Constructing CTSMV Instruments for History Teachers.”

ABSTRACT Using PISA 2022 data from 67 countries, this study conducts a two-stage meta-analysis to examine the language-related reading divide and its multilevel ecological moderations. The findings show that students from non-test-language backgrounds experience a significant reading disadvantage, and that the magnitude of this disadvantage varies substantially across countries. By integrating macro- and micro-level factors within a single educational ecological framework, the study identifies two opposing moderating pathways. At the micro level, multilingual and multicultural teacher training shows a significant narrowing effect and is associated with a smaller reading disadvantage among linguistically disadvantaged students. At the macro level, the Human Development Index exhibits a widening effect, suggesting that language-related disadvantage may be more pronounced in more developed national contexts. These results indicate that national development alone does not automatically translate into stronger protection for linguistically disadvantaged groups. Instead, reducing the language-related reading divide requires both targeted teacher training and more inclusive institutional arrangements. Overall, the study contributes to research on language diversity and educational equity by highlighting how multilevel educational environments shape students’ reading outcomes.

Abstract We investigate a notion related to Chafe’s One New Idea Constraint (ONICON), namely that clauses only contain one new idea at a time. We investigate both the nominal and the verbal domain in discourse production data from 16 diverse languages. We find no convincing evidence for a hard constraint on the number of new ideas per clause, nor for speakers actively managing information flow in discourse. The absence of an underlying ONICON is suggested by (1) the fact that NPs co-occurring in a single clause do not have more accessible referents, (2) that observed distributions of information can likewise be reproduced by a randomised mechanism, and (3) the absence of a presentational referent activation pattern. A more likely account of observed discourse production pattern is producers’ overall goal to be both contentful and coherent, the former implying the mentioning of a considerable number of entities and the latter implying predications about these entities. It is in this sense that there is audience design, rather than in an avoidance of information overcharge.

Prior research indicates that verbal reports of illness perceptions are associated with quality of life among individuals with interstitial lung disease (ILD). However, individuals with diverse language and verbal abilities may have difficulty providing verbal ratings of illness status. This study evaluated both verbal and visuo-graphic measures of perceived illness as predictors of illness-related quality of life and self-reported pulmonary symptoms among individuals with ILD. This was a cross-sectional study in a convenience sample of patients with ILD recruited from a clinic setting. Forty adults (63% female; mean age: 62.8; range: 32-85 years) with ILD completed standard medical evaluations of pulmonary function and exercise capacity. Study participants also completed self-report questionnaires (pulmonary symptoms, illness-related quality of life, and illness perceptions) and the lung coloring task (LCT), a standardized visuo-graphic measure of illness perceptions developed for this study. The primary approach to data analysis was multiple regression analysis predicting illness-related quality of life and self-rated symptoms from pulmonary function, illness perceptions, and LCT. Both forced vital capacity and illness perceptions were significant predictors of illness-related quality of life, accounting for approximately 55% of the variance. The symptom dimension of illness-related quality of life was predicted by forced vital capacity, illness perceptions, and LCT, accounting for approximately 40% of the variance. Illness perceptions are important for understanding illness-related quality of life among individuals with chronic lung disease. Visuo-graphic measures of illness perceptions may be especially useful for evaluating symptom experiences in chronic lung disease.

Ecosystem conservation and management: A theoretical framework using biocultural diversity.

The advantages of bilingualism on executive functioning have received intense attention over the past decade. However, characterizing what makes someone bilingual versus monolingual is complicated. Experiment 1 explores the impact of multilingualism and English-Mandarin language context priming on flanker task performance. Here, multilingualism is measured using the Multilingual Language Diversity (MLD) score. Experiment 2 investigates the influence of a familiar-unfamiliar language context priming (English-Welsh) on the flanker task performance. Experiments 1 and 2 found no significant differences between language contexts. However, the mean RT was similar between the control block and the mixed-language block in Experiment 1, but significantly different in Experiment 2. Only in Experiment 1 did higher MLD scores predict better performance on the flanker task, but only in mixed-language contexts (e.g., English and Mandarin) and not in single-language contexts (e.g., English or Mandarin only) While there were no significant language block differences in Experiment 2, the patterns of results were unexpectedly similar to Experiment 1, where the fastest RT was the control, followed by the mixed and single language blocks. This suggests that the effects of language contexts and the degree of multilingualism are limited to executive function. Additionally, while there were no effects of language familiarity on executive function, the pattern was similar for familiar and unfamiliar language contexts, and this warrants further investigation.

Strategies to Engage Diverse Population in Home-Based Palliative Care: A Qualitative Study in Hawaii.

Background. Personalized learning systems are expanding rapidly in higher education, but many recommendation pipelines still depend on structured indicators such as grades, attendance, task completion, and clickstream activity. In transnational educational settings, those indicators cannot adequately explain how students interpret disciplinary language, negotiate cultural expectations, or express learning difficulties. Purpose. This study develops a narrative-informed framework for personalized learning recommendations by integrating structured academic data and student-authored digital narratives within a multimodal learning analytics perspective. Method. The manuscript is positioned as a design science and framework-development study rather than a completed quantitative experiment. It synthesizes recent literature on learning analytics, educational recommender systems, multilingual education, natural language processing, and human-centred AI in education to specify a technical workflow for narrative preprocessing, multimodal fusion, learner-state modelling, recommendation generation, and evaluation. The small learner records presented in tables are synthetic examples used only to illustrate the data architecture. Results. The main output is a technically explicit and theoretically grounded framework that explains how narrative text can be anonymized, segmented, normalized, encoded into multilingual embeddings, combined with numerical learner indicators through early and late fusion, and evaluated using both predictive and recommender metrics. The framework also operationalizes transnational variables, including language diversity, culturally indirect participation, and contextual adaptation needs. Conclusion. Digital narratives can enrich learner profiling, improve contextual sensitivity, and strengthen culturally responsive recommendations. The study contributes a coherent blueprint for future empirical implementation in multilingual and transnational learning environments

Aim. The present research explores the cognitive principles that underlie the selection of negation types in Jordanian Arabic (JA) while expressing futurity. Methods. Data was collected by asking speakers of Jordanian Arabic to perform an acceptability judgement task by rating according to a 5-point Likert scale the three negation types: preverbal (*mā*), postverbal (*…š*), and bipartite (*mā…š*). Results. Findings have shown a strong preference for preverbal negation with the proclitic mā, as opposed to the discontinuous (bipartite) negation construction with mā …… š which was consistently rated as the least acceptable, particularly when combined with the lexical future marker bedd- Postverbal negation with the enclitic …š was slightly less acceptable, while the discontinuous (bipartite) negation construction with mā …… š was least favored by the participants. Analysis of participants’ responses indicated that this preference is driven by verbal and cognitive economy; speakers favored forms that offer better processing efficiency and articulatory simplicity. This, in turn, challenges the phonologically based explanation for negation preference, as the preference forpreverbal negation was high even with labial initial verbs. Conclusions. The study provides evidence that the choice among negation types in Jordanian Arabic (JA) is neither arbitrary nor purely pragmatic; instead, it is governed by cognitive constraints of economy and processing efficiency. A significant preference for preverbal negation over bipartite negation was identified. It was established that the use of preverbal structures ensures early polarity identification and reduces working memory load, whereas bipartite structures increase processing complexity. These findings align with the Uniform Information Density (UID) hypothesis, confirming the influence of cognitive principles on syntactic choice across typologically diverse languages. The research demonstrates that processing efficiency principles are fundamental to speech production and take priority over phonological constraints. An exception occurs when bipartite negation is utilized to fulfill the pragmatic function of emphasis. Regarding language change, it is substantiated that preverbal negation is more resistant to transformation, while more complex bipartite structures are vulnerable to reduction. The author concludes that cycles of language change reflect underlying cognitive preferences for structures optimal for cognitive processing.

Multilingual learning is key in natural language processing, but is challenged by the transfer–interference trade-off, where positive transfer benefits certain languages, while negative interference affects others. Prior methods, including linguistic-based and embedding-based language clustering, have attempted to address this; yet, they remain constrained by their static design and lack of task-specific feedback. In this study, we propose a novel computational strategy inspired by molecular design that constructs molecules with targeted properties. Languages are modeled as nodes in an undirected graph, with edges representing the transfer strength. This language molecule is optimized via Reinforcement Learning to adjust edge connections and weights to enhance positive transfer and minimize interference, where graph clustering is applied, and clusters are then evaluated on the Named Entity Recognition and POS tagging tasks using 25 languages from the WikiANN dataset and 12 typologically diverse languages from the UDPOS dataset. Compared to linguistic and embedding-based language clustering baselines, our method yields substantial improvements, especially for low-resource languages, with some showing over 35% increase in F1 score, while high-resource languages benefit moderately, confirming reduced transfer–interference trade-off. Our atom–language model offers a novel path for multilingual learning, inspired by molecular principles from physical sciences.

With the rapid development of public digital cultural resources, the lack of cross-lingual information retrieval (CLIR) services catering to multilingual users in practical applications has created significant language barriers. This hinders the promotion of public digital culture and results in the underutilization of relevant resources. To address this need, this paper constructs M-APE, a shared semantic model that operates without reliance on parallel corpora. Through a three-step process comprising the generation, fine-tuning, and optimization of a shared semantic space, M-APE establishes a common semantic framework for diverse languages. The model utilizes a Chinese semantic space, transferred and trained on authentic public cultural corpora, as its input. Evaluation based on bilingual dictionary induction quality demonstrates that M-APE significantly enhances semantic sharing performance between Chinese and Indo-European languages, represented here by English and French, achieving an average cross-family transformation accuracy of 56.6%. Furthermore, focusing on the CLIR needs of multilingual users within China’s public cultural engineering projects, this study develops a Chinese-English-French cross-lingual information retrieval framework by integrating M-APE into public cultural domain tasks. Experimental results indicate that the proposed method achieves superior cross-lingual retrieval performance in terms of average metrics.

Electromagnetic Articulography (EMA), an instrument to capture high-precision 3D movement data of the tongue, lips, and jaw, has been widely employed in speech production research, L2 instruction, and clinical rehabilitation. This study introduces the SAIT-EMA database, a Mandarin speech corpus that integrates articulatory and acoustic data from both native speakers of Mandarin and second language learners from diverse first-language backgrounds. The database comprises data from 18 participants (9 male and 9 female), including six L2 learners from Vietnam, Peru, Mexico, and Russia, and twelve native Mandarin speakers as a control group (two highly proficient speakers and ten general native speakers). Data were collected using the Carstens AG501 EMA system, with nine sensors attached to articulators: tongue tip (TT), tongue blade (TB), tongue dorsum (TD), lower incisors (LI), upper lip (UL), lower lip (LL), nasal bridge (NR), and bilateral mastoids (LE and RE). Articulatory data were sampled at 250 Hz, with synchronized audio recorded at 48 kHz, resulting in motion data in.pos format and standard.wav audio files. The speech materials were selected from a Mandarin interlanguage speech corpus, covering a broad range of phonetic phenomena including segments, tones, tone sandhi, and focus prosody-ensuring coverage of key issues in L2 phonetic research. The SAIT-EMA database offers the resource specifically targeting articulatory patterns in Mandarin L2 learners from diverse language backgrounds. It provides a valuable foundation for testing theoretical hypotheses in L2 acquisition and supports the development of intelligent computer aided pronunciation training technologies.

This quasi-experimental study investigates the impact of structured writing exercises on the sentence-writing abilities of grade 6 pupils at Jose P. Laurel Elementary School in General Santos City, Philippines. Participants were divided into two groups: an experimental group receiving targeted writing exercises and a control group engaged in regular classroom activities. Both groups were assessed using a standardized writing rubric before (pre-test) and after (post-test) the intervention period to evaluate their performance across four language structure taxonomies: morphology, syntax, lexicon, and orthography. The findings reveal that both the experimental and control groups exhibited initial poor performance in all four language structure taxonomies. However, only the experimental group demonstrated significant improvement in orthographic skills, suggesting that the structured writing exercises had a positive impact on this specific area. No significant differences were observed in morphological, syntactic, or lexical skills between the groups or within the control group. These results indicate that while the implemented writing exercises may have some positive effects on specific aspects of writing, further refinement of the intervention is necessary to address the diverse language learning needs of sixth-grade learners effectively. Future research may explore different types of writing exercises and their impact on various aspects of sentence-writing abilities.

ABSTRACT As Spanish for Specific Purposes (SSP) continues to grow at U.S. universities, the need for research on its teaching and learning is expanding. As more colleges provide Open Educational Resources (OER), SSP instructors have created, adopted, or adapted OER. This study examines OER’s role in SSP and investigates the extent to which SSP instructors use or refrain from using OER. The researchers surveyed 81 SSP instructors at U.S. institutions and then interviewed 11 of them in individual or focus group settings. Using constructivist grounded theory, the team coded transcripts line by line and grouped codes into categories related to OER in SSP: SSP Course Characteristics, Types of OER, Generative Artificial Intelligence (AI) and OER, Confusion about OER Terminology, Collaboration, and Benefits of OER and Barriers to OER or disadvantages of OER in SSP. Additional topics that emerged included community engagement, diverse language proficiency, and cultural competence related to OER used in SSP courses. Participants indicated a need for higher-quality, field-specific OER and greater institutional support for creating, adapting, and revising materials. This study shows that college-level SSP instructors require more evidence-based research and collaboration to develop and effectively implement OER in SSP courses.

Automated interpretation of Markush structures widely used in pharmaceutical patents to claim large families of related compounds remains challenging due to non-machine-readable structure images, variable R-groups, dependency rules, scaffold diversity, and heterogeneous claim language. Challenges include attachment points and stereochemistry, nested/conditional dependencies, and inconsistent drafting conventions that hinder faithful enumeration. Early rule-based cheminformatics systems parsed claims and mapped Markush representations into searchable formats, but struggled with nested dependencies, cross-references, and multimodal (text + image) descriptions. More recently, artificial intelligence (AI) methods have been introduced including language-based tools, vision-based tools, and multimodal or hybrid tools. Language-based tools increasingly use large language models (LLMs) and natural language processing (NLP) capabilities to extract variable definitions, constraints, and dependency graphs from claim text; vision systems translate structure depictions into machine-readable formats (e.g., SMILES, CXSMILES); multimodal or hybrid pipelines align both for end-to-end interpretation. Emerging datasets support these efforts, though licensing, family-wise leakage, and standardized splits remain inconsistent. This narrative review synthesizes tools, datasets, and evaluation practices for AI-assisted Markush interpretation, identifies persistent failure modes, and maps open legal questions (sufficiency, enablement, enforceability). We outline priorities for the field; transparent benchmarks with family-aware splits, interpretable constraint handling, and workflows aligned with U.S. Patent Office practice, near-term use is decision support, not legal advice.