Search Results (116)

English for Academic Purposes (EAP) courses primarily aim to cultivate academic communication, yet English-only norms and exam-oriented histories often discourage bilingual participation. This qualitative study traced Chinese graduate students’ language-belief development over one semester in a graduate EAP course and examined how the instructor mediated that process. Data included two rounds of open-ended surveys in two intact classes (N = 40), two interview rounds and end-of-semester reflections from ten purposively selected focus students (n = 10), and video-recorded classroom observations of 12 lessons. Findings show that the students increasingly legitimized bilingual participation and reframed English learning from test preparation toward academic communication. Beliefs nevertheless remained layered. Many still upheld an English-only ideal, treated English as the default language, and positioned the first language (L1) mainly as support when second language (L2) expression became difficult. Endorsement also exceeded uptake, with L1 use treated as a compensatory fallback rather than a co-equal academic resource. Instructor policy, conceptual framing, and interactional modeling reduced anxiety around bilingual moves and sometimes supported greater willingness to attempt more English, which identifies mechanisms for bilingual-aware EAP pedagogy in monolingual-leaning EFL contexts. Full article

Down syndrome (DS) is associated with persistent language impairments that extend beyond early childhood, yet evidence from agglutinative languages remains limited. While morphosyntactic weaknesses have been well-documented in Indo-European languages, less is known about how such difficulties are manifested in Turkish, a language in which grammatical relations are primarily marked through morphology. In addition, short-term memory (STM) limitations, particularly in verbal domains, are characteristic of DS and may contribute to language outcomes. This study examined the interaction between morphosyntax and STM in Turkish-speaking children and adolescents with DS. A cross-sectional observational design was employed, including 12 monolingual Turkish-speaking participants with DS (aged 6;7–15;11) and 10 TD peers matched on nonverbal mental age. Participants completed standardized assessments of syntax and morphology, spontaneous language sampling, and STM tasks assessing verbal and visual memory. Children with DS performed significantly below controls on syntactic comprehension and production as well as morphological measures, with larger effects observed for syntax. Noun morphology was less accurate than verb morphology, likely reflecting increased morphophonological complexity. Regression analyses indicated that auditory digit span predicted sentence comprehension, whereas nonword repetition predicted morphological production indexed by mean length of utterance in morphemes. Substantial inter-individual variability was observed within the DS group. These findings suggest that morphosyntactic outcomes in Turkish-speaking children with DS are closely linked to verbal STM capacities and vary considerably across individuals, underscoring the importance of integrated assessment and individualized intervention planning. Future research with larger samples is warranted to confirm and extend these preliminary findings. Findings should be interpreted cautiously due to the limited sample size and are presented as preliminary descriptive evidence. This study provides initial data on Turkish-speaking individuals with Down syndrome. Full article

This study explores the perspectives of monolingual English-speaking parents whose children are enrolled in elementary (Grades 1–5) French immersion (FI) in New Brunswick, Canada, where FI students learn mathematics in French. Using poetic inquiry within a feminist postmodern framework, we analyzed interview data from three parents to examine how they conceptualize the relationship between language and mathematics, and how these conceptualizations shape the ways they support their children’s mathematics learning. The resulting research poems reveal tensions in participants’ views of mathematics and language. For example, mathematics was at times positioned as detachable from language, although language was simultaneously described as a potential barrier to mathematical success. In turn, parental involvement was characterized by support toward monitoring linguistic markers, relearning pedagogical methods, and rehearsing procedures. By centring parents’ perspectives, this study contributes to research on multilingual mathematics education by illustrating how parental conceptualizations may play a role in shaping mathematics practices across home and school spaces. Methodologically, the study suggests that research poetry has analytic potential for surfacing tensions in parental sense-making that may remain overlooked in more conventional qualitative analyses. This study points to a need for resources and communication practices that support dialogue between schools and families about the relationship between language and mathematics in FI contexts. Full article

This study explores how adolescent learners conceptualize the emotion of love in their first language (Spanish) and in English as a foreign language (EFL), comparing monolingual Spanish speakers and Spanish–Arabic bilinguals. A total of 66 participants (33 per group), all with A2 proficiency in English, completed a semantic fluency task in both Spanish and English, producing as many words as possible in relation to the prompts Amor and Love. The data were analyzed using graph theory to capture the organization of participants’ emotion lexicons. The results show that love is a highly productive and cohesive semantic field, eliciting significantly more responses in L1 than in L2, for both Spanish-only (t = −8.866, p < 0.001) and Spanish–Arabic (W = 101.0, p = 0.001) participants. The differences between the two learner cohorts were not significant in Spanish nor in English. The results from the graph analyses revealed that learners displayed rich and strongly connected networks in Spanish, with learners with a migration origin showing slightly more fragmented networks. In English, both groups performed similarly, with responses probably mediated by L1 translation equivalents and metaphorical associations (e.g., heart, flower, and red). The findings suggest that emotional lexicons are better developed and more efficiently organized in the L1, whereas FL representations are shaped by proficiency, context of learning, and reliance on L1 conceptual structures. This study contributes novel insights into bilingual and heritage learners’ emotional conceptualization and highlights the value of graph analysis for examining the structure of emotion words. Full article

Background/Objectives: The “tip of the tongue” (TOT) and “feeling of knowing” (FOK) phenomena were cognitive experiences that notably affected word retrieval, particularly among older adults. The study aimed to investigate the influences of age of acquisition (AoA), vocabulary size, and verbal fluency on the frequency and nature of TOT and FOK occurrences as individuals aged. Methods: A behavioral experiment was conducted based on the two-step word retrieval framework established by Gollan and Brown in 2006. Early and late acquisition words were utilized to induce tip-of-the-tongue phenomena and the feeling of knowing. Additionally, vocabulary and verbal fluency tests were administered. Sixty monolingual older adults participated in the study (35 female, 25 male; mean age: 77.66 years). Mixed-effects linear regressions had been used to analyze the data. Results: The logistic regression analysis identified age of acquisition as the most significant predictor of TOT and FOK experiences (p < 0.0001), highlighting that earlier vocabulary acquisition enhanced retrieval efficiency. Notable interactions between vocabulary size and verbal fluency illustrated that increased lexical knowledge diminished reliance on age of acquisition for successful retrieval. Conclusions: The findings underscore the importance of early vocabulary acquisition as a protective factor against cognitive decline in older adults, emphasizing the necessity for interventions aimed at enhancing vocabulary and fluency. This study contributed valuable insights into the cognitive mechanisms underlying language retrieval and suggested that fostering rich linguistic environments throughout life could facilitate better cognitive health in aging populations. Full article

Large Language Models (LLMs) excel in translation, among other things, demonstrating competitive performance for many language pairs in zero- and few-shot settings. But unlike dedicated neural machine translation models, LLMs are not trained on any translation-related objective. What explains their remarkable translation abilities? Are these abilities grounded in “incidental bilingualism” in training data? Does instruction tuning contribute to it? Are LLMs capable of aligning and leveraging semantically identical or similar monolingual contents from different corners of the internet that are unlikely to fit in a single context window? I offer some reflections on this topic, informed by recent studies and growing user experience. My working hypothesis is that LLMs’ translation abilities originate in two different types of pre-training data that may be internalized by the models in different ways: Local and Global. “Local learning” makes use of bilingual signals present within a single training context window (e.g., an English sentence soon followed by its Chinese translation in the training data). “Global learning,” in contrast, capitalizes on mining semantically related monolingual contents that are spread out over the LLMs’ pre-training data. The key to explaining the origins of LLMs’ translation capabilities is a continuous iteration between Local and Global learning, which is a natural and helpful consequence of batch training. I discuss the prospects for testing the “duality hypothesis” empirically and its implications for reconceptualizing translation, human and machine, in the age of deep learning. Full article

This study introduces a Romanian–English bilingual corpus and a fine-tuned cross-lingual embedding framework aimed at improving retrieval performance in Retrieval-Augmented Generation (RAG) systems. The dataset integrates over 130,000 unstructured question–answer pairs derived from SQuAD and 9750 Romanian-generated questions linked to governmental tabular data, subsequently translated bidirectionally to build parallel Romanian–English resources. Multiple state-of-the-art embedding models, including multilingual-e5, Jina-v3, and the Qwen3-Embeddings family, were systematically evaluated on both text and tabular inputs across four language directions (eng-eng, ro-ro, eng-ro, ro-eng). The results show that while multilingual-e5-large achieved the strongest monolingual retrieval performance, Qwen3-Embedding-4B provided the best overall balance across languages and modalities. Fine-tuning this model using Low-Rank Adaptation (LoRA) and InfoNCE loss improved its Mean Reciprocal Rank (MRR) from 0.4496 to 0.4872 (+8.36%), with the largest gains observed in cross-lingual retrieval tasks. The research highlights persistent challenges in structured (tabular) data retrieval due to dataset imbalance and outlines future directions including dataset expansion, translation refinement, and instruction-based fine-tuning. Overall, this work contributes new bilingual analyses and methodological insights for advancing embedding-based retrieval in low-resource and multimodal contexts. Full article

Bimodal emotion recognition based on audio and text is widely adopted in video-constrained real-world applications such as call centers and voice assistants. However, existing systems suffer from limited cross-domain generalization and monolingual bias. To address these limitations, a cross-lingual bimodal emotion recognition method is proposed, integrating Mamba-based temporal encoders for audio (Wav2Vec2.0) and text (Jina-v3) with a Transformer-based cross-modal fusion architecture (BiFormer). Three corpus-adaptive augmentation strategies are introduced: (1) Stacked Data Sampling, in which short utterances are concatenated to stabilize sequence length; (2) Label Smoothing Generation based on Large Language Model, where the Qwen3-4B model is prompted to detect subtle emotional cues missed by annotators, producing soft labels that reflect latent emotional co-occurrences; and (3) Text-to-Utterance Generation, in which emotionally labeled utterances are generated by ChatGPT-5 and synthesized into speech using the DIA-TTS model, enabling controlled creation of affective audio–text pairs without human annotation. BiFormer is trained jointly on the English Multimodal EmotionLines Dataset and the Russian Emotional Speech Dialogs corpus, enabling cross-lingual transfer without parallel data. Experimental results show that the optimal data augmentation strategy is corpus-dependent: Stacked Data Sampling achieves the best performance on short, noisy English utterances, while Label Smoothing Generation based on Large Language Model better captures nuanced emotional expressions in longer Russian utterances. Text-to-Utterance Generation does not yield a measurable gain due to current limitations in expressive speech synthesis. When combined, the two best performing strategies produce complementary improvements, establishing new state-of-the-art performance in both monolingual and cross-lingual settings. Full article

Discourse markers have been extensively studied in spoken languages from different perspectives, covering monolingual, contrastive, and translation studies. However, research on these items remains limited for signed languages, with only a handful of scattered publications. Following a corpus-based approach, this paper aims to investigate discourse markers in French Belgian Sign Language (LSFB), including their types, functions, and translation/s into written French. An 18 min sample of three dialogues and six signers was analyzed using a two-level independent taxonomy (domain and function) previously applied to spoken and signed data. Overall, 251 discourse markers were identified in the LSFB sample. They can be manual, nonmanual, or a combination of both, the latter type being the most frequent. In contrast to the previous literature, discourse markers cannot be spatial in LSFB. Regarding their functional spectrum, most discourse markers belong to the sequential domain (i.e., they are mostly used to structure discourse) and express ‘addition’ (i.e., providing more information) or ‘monitoring’ (i.e., keeping control over one’s turn or over the interaction). When examining the translation of DMs, most are either omitted or substituted by other non-discourse marking items in the target texts. Although these results are generally similar to previous studies on DMs in spoken languages, more research on these items in other signed languages is needed to obtain a precise overview of their role in human communication. Full article

This single case study examines the implementation of a co-designed fifth-grade science unit enhanced by using Virtual Reality (VR) and integrating translingual and transmodal scaffolding strategies to support students’ participation and quality of talk during scientific sensemaking. The co-designed science unit covered physical and chemical changes as part of the fifth-grade science curriculum. The research involves a fifth-grade science teacher and her class of 22 students comprising multilingual learners (ML) and English monolingual learners (EML). This study examines the learning experience of 3 student pairs grouped as ML-ML, EML-ML and EML-EML. Using content analysis, we analyzed 911 min of video data on the six students’ learning in this unit. The results indicate that when the teacher used translingual and transmodal scaffolding strategies introduced during the co-design process, equal participation across MLs and EMLs was observed. The VR pair programming worked well for student pairs in increasing active participation regardless of the pairing, although active participation did not necessarily lead to high quality science talk. Findings of this study provide implications and recommendations for leveraging the scaffolding from teachers, materials, and VR pair programing activity to support the equal participation and quality of talk among all learners during scientific sensemaking. Full article

Background: Children with neurodevelopmental disorders (NDDs) commonly experience executive function (EF) impairments that impact daily life and academics. While bilingualism has generally been associated with cognitive advantages in typically developing (TD) children, its relationship with EF in children with NDDs remains unclear and represents a critical knowledge gap for families and clinicians considering bilingual exposure in these populations. Methods: For this scoping review, we searched PubMed, ProQuest, CogNet, PsycINFO, Scopus, ERIC, Embase, CINAHL, Linguistics Abstracts Online, and Google Scholar for studies published between database inception and December 2024, without language restrictions. We included quantitative, qualitative, and mixed-methods studies that (i) involved participants aged 4–12 years with diagnosed NDDs; (ii) examined children with bilingual language exposure; (iii) employed validated instruments for measuring cognitive or executive function; (iv) presented original empirical findings; and (v) were published in English. We excluded studies lacking comparisons between groups and longitudinal studies. Data on study characteristics, participants, EF assessments, and main findings were extracted. This study is registered with OSF Registries. Findings: Fifteen cross-sectional studies met the inclusion criteria, all of which focused exclusively on children with autism spectrum disorder (ASD), with no studies examining other NDDs. The studies involved 982 children with ASD (463 monolingual; 404 bilingual) and 644 TD children. Most studies (n = 11) revealed that, compared with monolingual children with ASD, bilingual children with ASD demonstrated advantages in working memory, cognitive flexibility, and inhibitory control on performance-based tasks. However, findings were inconsistent for spatial inhibition tasks, and parent-reported measures sometimes failed to detect bilingual-related differences. Interpretation: Bilingualism is associated with specific EF benefits for children with ASD, adding to evidence that questions longstanding concerns about the negative impacts of bilingual exposure in NDD populations. The evidence suggests that bilingual exposure could potentially serve as a complementary approach to traditional interventions for addressing EF impairments in children with ASD, although this evidence is limited to cross-sectional designs and requires further studies. However, the exclusive focus on ASD limits generalisability across the broader spectrum of NDDs. Further research is needed across diverse NDD populations employing comprehensive, multi-method EF assessments that combine performance-based tasks with parent-reported measures to better inform parenting, clinical, and educational practices. Full article

The success of current entity alignment (EA) tasks largely depends on the supervision information provided by labeled data. Considering the cost of labeled data, most supervised methods are challenging to apply in practical scenarios. Therefore, an increasing number of works based on contrastive learning, active learning, or other deep learning techniques have been developed, to solve the performance bottleneck caused by the lack of labeled data. However, existing unsupervised EA methods still face certain limitations; either their modeling complexity is high or they fail to balance the effectiveness and practicality of alignment. To overcome these issues, we propose a learnable convolutional attention network for unsupervised entity alignment, named LCA-UEA. Specifically, LCA-UEA performs convolution operations before the attention mechanism, ensuring the acquisition of structural information and avoiding the superposition of redundant information. Then, to efficiently filter out invalid neighborhood information of aligned entities, LCA-UEA designs a relation structure reconstruction method based on potential matching relations, thereby enhancing the usability and scalability of the EA method. Notably, a similarity function based on consistency is proposed to better measure the similarity of candidate entity pairs. Finally, we conducted extensive experiments on three datasets of different sizes and types (cross-lingual and monolingual) to verify the superiority of LCA-UEA. Experimental results demonstrate that LCA-UEA significantly improved alignment accuracy, outperforming 25 supervised or unsupervised methods, and improving by 6.4% in Hits@1 over the best baseline in the best case. Full article

This eye-tracking study examined how heritage speakers of Spanish process gender agreement morphology at a distance, focusing on the activation of the gender feature during sentence processing. Previous work is conceptually replicated and further extended by assessing (1) whether reduced sensitivity to gender agreement mismatches when another word intervenes between the head noun and its modifying adjective stems from weakened gender feature activation, (2) whether a gender-marked determiner enhances this activation, and (3) whether Age of Onset of Bilingualism (AOB) plays a role in this activation. Fifty-three English-dominant heritage speakers of Spanish and a comparison group of 32 Spanish-dominant monolingually raised speakers read sentences with and without gender agreement mismatches while their eye movements were monitored. Sentences contained mismatches in adjectives modified by the intensifier “muy” under two conditions: a No Cue condition (e.g., árboles muy altos/*altas) and a Cue condition with a gender-marked determiner (e.g., unos árboles muy altos/*altas). Statistical modeling of the eye-tracking data suggests similar effects for both groups in the No Cue condition, but AOB and proficiency modulated sensitivity for heritage speakers with a later AOB (4–6). Gender cues on the determiner (Cue condition) impacted the time course of agreement processing for all groups, the total time spent reading mismatches for all heritage speakers as a function of proficiency, and the rereading time for heritage speakers with a later AOB (4–9). We consider the role of Age of Onset of Bilingualism (AOB) and proficiency in morphosyntactic processing, feature retrieval, and cue facilitation in heritage language processing. Full article

Cultural heritage institutions, such as museums and galleries, today face the challenge of managing an increasing volume of unsolicited visitor feedback generated across online platforms. This study offers a practical and scalable methodology that transforms 5856 multilingual Google reviews from 59 globally ranked museums and galleries into actionable insights through sentiment analysis, correlation diagnostics, and guided Latent Dirichlet Allocation. By addressing the limitations of prior research, such as outdated datasets, monolingual bias, and narrow geographical focus, the authors analyze a current and diverse set of recent reviews to capture a timely and globally relevant perspective on visitor experiences. The adopted guided LDA model identifies 12 key topics, reflecting both operational issues and emotional responses. The results indicate that while visitors generally express overwhelmingly positive sentiments, dissatisfaction tends to be concentrated in specific service areas. Correlation analysis reveals that longer, emotionally rich reviews are more likely to convey stronger sentiment and receive peer endorsement, highlighting their diagnostic significance. From a practical perspective, the methodology empowers professionals to prioritize improvements based on data-driven insights. By integrating quantitative metrics with qualitative topics, this study supports operational decision-making and cultivates a more empathetic and responsive data management mindset for museums. The reproducible and adaptable nature of the pipeline makes it suitable for cultural institutions of various sizes and resources. Ultimately, this work contributes to the field of cultural informatics by bridging computational precision with humanistic inquiry. That is, it illustrates how intelligent analysis of visitor reviews can lead to a more personalized, inclusive, and strategic museum experience. Full article

This study presents the first systematic empirical comparison of neural architectures for Universal Dependencies (UD) parsing in Old English, thus addressing central questions in computational historical linguistics and low-resource language processing. We evaluate three approaches—a baseline spaCy pipeline, a pipeline with a pretrained tok2vec component, and a MobileBERT transformer-based model—across datasets ranging from 1000 to 20,000 words. Our results demonstrate that the pretrained tok2vec model consistently outperforms alternatives, because it achieves 83.24% UAS and 74.23% LAS with the largest dataset, whereas the transformer-based approach substantially underperforms despite higher computational costs. Performance analysis reveals that basic tagging tasks reach 85–90% accuracy, while dependency parsing achieves approximately 75% accuracy. We identify critical scaling thresholds, with substantial improvements occurring between 1000 and 5000 words and diminishing returns beyond 10,000 words, which provides insights into scaling laws for historical languages. Technical analysis reveals that the poor performance of the transformer stems from parameter-to-data ratio mismatches (1250:1) and the unique orthographic and morphological characteristics of Old English. These findings defy assumptions about transformer superiority in low-resource scenarios and establish evidence-based guidelines for researchers working with historical languages. The broader significance of this study extends to enabling an automated analysis of three million words of extant Old English texts and providing a framework for optimal architecture selection in data-constrained environments. Our results suggest that medium-complexity architectures with monolingual pretraining offer superior cost–benefit trade-offs compared to complex transformer models for historical language processing. Full article