Search Results (1,311)

Harmful language in conversational AI poses distinct safety and governance challenges, as Large Language Model (LLM) chatbots interact in private, one-to-one settings. Understanding the types of harm and their temporal concentration is crucial for responsible deployment and time-aware moderation. This study investigates the types and diurnal dynamics of harmful speech, comparing patterns between play-oriented chit-chat and task-oriented LLM services.We analyze two large-scale, real-world English corpora: a chit-chat service (SimSimi; 8.7 M utterances) and an LLM service (WildChat; 610 K utterances). Using the Perspective API for multi-label classification (Toxicity, Profanity, Insult, Identity Attack, Threat), we estimate the incidence of harm categories and compare their distribution across five dayparts. Our analysis shows that harmful speech is significantly more prevalent in the chit-chat context than in the LLM service. Across both platforms, Toxicity and Profanity are the dominant categories. Temporally, harmful speech concentrates most frequently during the dawn daypart. We contribute an empirical baseline on how harm varies by chatbot modality and time of day, offering practical guidance for designing dynamic, platform-specific moderation policies. Full article

The family serves as a critical domain for ethnic minority children to acquire their ethnic language. It plays a vital role in fostering multilingual competence and sustaining language diversity. Therefore, the family language planning (FLP) of ethnic minorities in China has attracted scholarly attention in recent years. However, there is relatively little research on the FLP of the Yi ethnic group. The present study investigated the FLP of the Yi (Nuosu) families in Liangshan Yi Autonomous Prefecture (Liangshan) through questionnaires, interviews, and ethnographic observations. The aim was to elucidate the language life within Nuosu families in the context of recent policy modifications regarding the instruction of Putonghua, the Nuosu language, and English. The results indicate that Nuosu families pay greater attention to Putonghua and English, and the Nuosu language practice in the family domain is decreasing across generations. There are signs of language shift from the Sichuan dialect to Putonghua in urban Nuosu families and from the Nuosu language to Putonghua in rural Nuosu families. The Nuosu parents adopt divergent language management strategies in response to the new language education policies. The findings contribute to our understanding of the FLP of a less-discussed ethnic group, as well as the interaction between FLP and school education policies. This study also provides a unique case in the field of multilingual studies, both domestically and internationally. Full article

Gold mining has played a central role in shaping Peruvian society from pre-Inca civilizations to the present. However, existing literature offers fragmented perspectives, often focusing on isolated themes such as metallurgy, colonial mercury use, or environmental degradation, without integrating these across time and territory. This review addresses that gap by offering a place-based synthesis that combines archaeological, historical, legal, environmental, and comparative insights. Drawing on both Spanish-language sources and international literature, the paper reconstructs Peru’s gold mining trajectory through five historical phases—pre-Inca, Inca, colonial, republican, and contemporary—highlighting continuities and ruptures in governance, labor systems, and environmental impacts. The analysis reveals persistent challenges in Peru’s gold sector, including informality, mercury pollution, and weak institutional capacity. Compared to other mining economies such as Chile, Ghana, and South Africa, Peru exhibits greater fragmentation and limited integration of mining into national development strategies. The review also explores the role of gold in the global energy transition, emphasizing its relevance in clean technologies and green finance, and identifies policy gaps that hinder Peru’s alignment with sustainability goals. By bridging linguistic and disciplinary divides, this synthesis contributes to a more inclusive historiography of extractive industries and underscores the need for interdisciplinary approaches to mining governance. Ultimately, the paper calls for a reimagining of Peru’s gold sector, one that prioritizes environmental justice, social equity, and long-term resilience. Full article

This study investigates the developmental trajectories of transcription and oral language skills in kindergarten students over the course of an academic year, with a focus on the influence of executive functions (EF) and home literacy practices (HLP). Hierarchical linear modeling (HLM) analyses revealed significant growth in transcription skills, with both EF and independent home literacy practices positively influencing baseline transcription scores. The interaction between independent home literacy practices and formal literacy practices at home further enhanced transcription skill development. In contrast, oral language skills were not influenced by either HLP or EF. These results suggest that EF plays a more prominent role in transcription development than oral language skills in early childhood, especially in transparent orthographic systems. The findings highlight the importance of cognitive and environmental factors in early literacy development, suggesting implications for educational practices, particularly in fostering effective home literacy environments Full article

Individual behaviors play a crucial role in generating and mitigating plastic pollution. Understanding citizen knowledge and perceptions is therefore critical to inform effective public interventions. Surveys can provide this information, but they must use well-designed and culturally adapted tools to be reliable. We present the Italian cross-cultural adaptation of an Australian questionnaire identified via systematic review as meeting high-quality standards. The tool included 21 items (Likert-scale, multiple-choice, and open-ended). In accordance with literature indications, we performed forward and back translation and subsequent review by an Expert Committee, producing a pre-final version. A stratified sample of 43 citizens assessed clarity of each item and provided feedback, which guided further Expert Committee revision. Ten items showed comprehension problems, and seven of them were rephrased because they were confusing or redundant. Items with technical terms such as “bioplastics” and “biodegradable” proved challenging, leading to the addition of brief explanations in the introduction to the questionnaire. This process produced a rigorously developed, culturally appropriate instrument for assessing public understanding of plastic pollution in Italy. This standardized tool, if adapted in multiple languages, will enable international surveys and meta-analyses to guide global strategies. Psychometric validation is recommended before large-scale deployment of the tool. Full article

The so-called canonical clause, consisting of case-marked NPs and a final finite verb, has played a central role in discussions of Japanese for the past several decades. The current study explores the nature of such clauses in everyday Japanese conversation. Everyday conversation is considered the most fundamental form of language, and large-scale corpora of Japanese everyday conversation have only become available in recent years, enabling projects like ours. One key finding is that clauses ending with a finite verb are rare, challenging the centrality of the canonical clause in Japanese grammar. Instead, we observe that the verb is usually followed by additional elements that convey pragmatic information. This observation suggests that the canonical clause for Japanese speakers should also include these pragmatic elements. We have observed further that the relatively uncommon examples that do end with a finite verb often involve five frequent semantically light verbs. A preliminary study of one of these verbs, chigau ‘to differ’, reveals that, typically without overt NPs, it functions more like a particle than the verb of a clause. This further calls into question the idea that the canonical clause in Japanese ends with a finite verb. Full article

This study integrates complex systems theory and innovation ecosystem theory to develop a unified framework encompassing the innovation environment, innovation actors, and innovation networks. Using fsQCA and NCA, it examines the impact of cross-layer interactions and the coupling of multiple factors on collaborative performance. Empirical analysis in the field of natural language processing (NLP) demonstrates that no single factor is sufficient to serve as a necessary condition for achieving high-innovation collaboration performance. Innovation actors, as endogenous evolutionary drivers, play a central and catalytic role in the collaboration process. Moreover, under specific conditions, the relationship between the innovation environment and innovation networks exhibits a substitutive effect, with certain capabilities enabling this dynamic. This study extends the theoretical understanding of collaboration pathways within innovation ecosystems and offers practical recommendations for fostering innovation cooperation across different industries and organizations. It achieves this by constructing a “situational type–configuration path” matrix, decision tree, and innovation collaboration performance realization model. Full article

Dementia disrupts communication not only as a cognitive process but as a relational practice, leaving people living with dementia (PLwD) at risk of exclusion when language fragments. This study examines how communication closeness, the felt sense of being understood, emotionally attuned, and socially connected, might be supported through Research in and through Design (Ri&tD). Drawing on formative mixed-reality studies and a participatory co-design workshop with PLwD, caregivers, and stakeholders, we iteratively developed a series of playful artifacts culminating in Neighbourly, a tactile board game designed to support relational interaction through rule-based, multimodal play. Across this design genealogy, prototypes were treated as Living Counter-Maps: participatory mappings that made patterns of gesture, rhythm, shared attention, and material engagement visible and discussable. Through iterative interpretation and synthesis, the study identifies three guiding principles for designing for communication closeness: supporting co-regulation rather than correction, enabling multimodal reciprocity, and providing a shared material focus for joint agency. The paper consolidates these insights in the Living Counter-Maps Framework, which integrates counter-mapping and Ri&tD as a methodological approach for studying and designing relational communication in dementia care. Full article

Learning the host language is a crucial factor in the settlement of migrants and refugees in a new country. It offers opportunities, but can also generate exclusion, marginalization, and isolation, hindering the possibilities of participation and the creation of social networks. Host language classes, therefore, play a crucial role, fostering critical awareness that enables learners to act within their social and cultural context. This promotes agency, autonomy, and empowerment, transforming differences into productivity and fostering social justice. This article focuses on classes of Portuguese as a host language (HL) in northern Portugal, examining how these contexts shape learners’ relationship with the language beyond depoliticized or subaltern approaches. It discusses the main obstacles and difficulties in these educational settings, as well as mechanisms that could contribute to more democratic and effective practices. Drawing on interviews with teachers (n = 10), trainers (n = 4), volunteers (n = 8), and students (n = 20) involved in the HL learning process, the content analysis highlights how policies and pedagogical practices impact students and how they are interpreted by these actors, revealing their impact on processes of participation, belonging, and citizenship. The results indicate an emergent form of collective autonomy in the relationship among students, the host society, and teachers, which means that teaching practices encompass not only the development of communication skills but also the civic and political awareness of learners. Lastly, while the language teachers identified more practical barriers in these teaching and learning contexts, the students described emotional and sociocultural obstacles. Full article

The assessment of Chinese text readability plays a significant role in Chinese language education. Due to the intrinsic differences between alphabetic languages and Chinese character representations, the readability assessment becomes more challenging in terms of the language’s inherent complexity in vocabulary, syntax, and semantics. The article proposed the conceptual analogy between Chinese readability assessment and music’s rhythm and tempo patterns, in which the syntactic structures of the Chinese sentences could be transformed into an image. The Chinese Knowledge and Information Processing Tagger (CkipTagger) tool developed by Sinica-Taiwan is utilized to decompose the Chinese text into a set of tokens. These tokens are then refined through a user-defined token pool to retain meaningful units. An image with part-of-speech (POS) information will be generated by using the token versus syntax alignment. A discrete cosine transform (DCT) is then applied to extract the temporal characteristics of the text. Moreover, the study integrated four categories: linguistic features–type–token ratio, average sentence length, total word, and difficulty level of vocabulary for the readability assessment. Finally, these features were fed into the Support Vector Machine (SVM) network for the classifications. Furthermore, a bidirectional long short-term memory (Bi-LSTM) network is adopted for quantitative comparisons. In simulation, a total of 774 Chinese texts fitted with Taiwan Benchmarks for the Chinese Language were selected and graded by Chinese language experts, consisting of equal amounts of basic, intermediate, and advanced levels. The finding indicated the proposed POS with the linguistic features work well in the SVM network, and the performance matches with the more complex architectures like the Bi-LSTM network in Chinese readability assessments. Full article

As large language models (LLMs) improve in understanding and reasoning, they are increasingly used in privacy protection tasks such as de-identification, privacy-sensitive text generation, and entity obfuscation. However, these applications depend on an essential requirement: the accurate identification of personally identifiable information (PII). Compared with template-based PII that follows clear structural patterns, name-related PII depends much more on cultural and pragmatic context, which makes it harder for models to detect and raises higher privacy risks. Although recent studies begin to address this issue, existing work remains limited in language coverage, evaluation granularity, and the depth of error analysis. To address these gaps, this study proposes an error-driven framework that integrates diagnosis and intervention. Specifically, the framework introduces a method called Error-Driven Prompt (EDP), which transforms common failure patterns into executable prompting strategies. It further explores the integration of EDP with general advanced prompting techniques such as Chain-of-Thought (CoT), few-shot learning, and role-playing. In addition, the study constructed K-NameDiag, the first fine-grained evaluation benchmark for Korean name-related PII, which includes twelve culturally sensitive subtypes designed to examine model weaknesses in real-world contexts. The experimental results showed that EDP improved F1-scores in the range of 6 to 9 points across three widely used commercial LLMs, namely Claude Sonnet 4.5, GPT-5, and Gemini 2.5 Pro, while the Combined Enhanced Prompt (CEP), which integrates EDP with advanced prompting strategies, resulted in different shifts in precision and recall rather than consistent improvements. Further subtype-level analysis suggests that subtypes reliant on implicit cultural context remain resistant to correction, which shows the limitations of prompt engineering in addressing a model’s lack of internalized cultural knowledge. Full article

Many studies have shown that flow, psychological capital (PsyCap), anxiety, and academic efficacy play significant roles in EFL learning, yet little attention has been paid to how these positive and negative states jointly shape learners’ PsyCap. Grounded in the broaden-and-build theory, this study investigated how flow, a state of deep engagement and enjoyment in learning, affected EFL learners’ PsyCap. A total of 1611 EFL learners at the CEFR B1–B2 levels from six universities in China participated in the study. Data were collected using validated questionnaires developed for this study that measured flow, foreign language classroom anxiety (FLCA), academic efficacy, and PsyCap, and analyzed using structural equation modeling (SEM) in AMOS. The results revealed that flow had a significant direct positive effect on PsyCap (β = 0.648, p < 0.001). Academic efficacy significantly mediated this relationship (β = 0.059, p < 0.001), and a significant chain-mediated path was observed through FLCA and academic efficacy (β = 0.023, p < 0.001). The total effect of flow on PsyCap was 0.729 (p < 0.001). These findings provide new insights into educational practices that can effectively enhance EFL learners’ PsyCap and academic achievement by facilitating flow and reducing anxiety. Full article

Large Language Models (LLMs) can adopt various roles through prompt guidance, enabled by pretraining on diverse corpora and instruction-following alignment. Safety alignment mechanisms attempt to define a helpful, honest, and harmless assistant to guide their behavior. However, specific role settings can bypass these safeguards, inducing LLMs to respond to harmful queries. In this study, we identify the role settings that lead LLMs to generate such harmful responses, which contribute to more reliable LLMs. We design an automated jailbreak framework, RoleBreaker, that optimizes role-play prompts with representation analysis and adaptive search. Experiments on 7 open-source LLMs show that RoleBreaker achieves an average jailbreak success rate of 87.3% with 4.0 attempts, outperforming SOTA methods. Furthermore, by summarizing the jailbreak experiences and applying them to closed-source commercial models (GPT-4.1, GLM-4, Gemini-2.0), we achieve an average jailbreak success rate of 84.3% with 4.3 attempts. These results reveal vulnerabilities in current alignment mechanisms and demonstrate the transferability of our approach. Full article

Large Language Models (LLMs) are emerging technologies and a growing research trend in Artificial General Intelligence (AGI), which envisions a future where machines can think and learn like humans across a wide range of tasks. Information generated by LLMs is essentially the prediction of next tokens in Natural Language Processing (NLP) tasks. However, the generated content is always subject to issues of truthfulness and hallucinations. The information and knowledge integrity of LLM-generated content therefore remains subjective. Exploring recent literature on the integrity of LLMs in a systematic manner is both timely and essential. Moreover, ensuring the reliability of LLMs in real-world applications is critical. Various approaches have been explored to promote information and knowledge integrity in LLMs, including adversarial training, data augmentation, and calibration methods. However, beyond these techniques, other strategies also contribute to maintaining knowledge integrity. This paper specifically focuses on three such approaches: knowledge distillation, semantic integrity, and provenance tracking, which play essential roles in ensuring that LLMs generate accurate, consistent, and trustworthy information. Knowledge distillation enhances model efficiency by transferring knowledge from larger models to smaller ones while preserving essential learning without compromising knowledge integrity. This reduces hallucinations. Semantic integrity safeguards consistency and strengthens the robustness of generated outputs. It is concurrently checking the meaningfulness of the outputs with the context. Provenance tracking improves transparency and trustworthiness through mechanisms such as data lineage and explainability, thereby ensuring the credibility of the LLM-generated responses. This review suggests that knowledge distillation, semantic integrity, and provenance tracking can enhance the reliability of LLM outputs, with prior studies reporting reductions in hallucination rates, improvements in robustness, and gains in factual consistency. Full article

Background and Objectives: Breast Implant-Associated Anaplastic Large Cell Lymphoma (BIA-ALCL) is a T-cell lymphoma that has shown an interest in the medical community in recent years. Given its emerging clinical relevance, accurate imaging plays a crucial role in diagnosis, staging, and follow-up. This systematic review aims to evaluate the role of Fluorine-18 Fluorodeoxyglucose positron emission tomography/computed tomography (18F-FDG PET/CT) in the staging and follow-up of patients with BIA-ALCL, focusing on its diagnostic accuracy and clinical impact. Materials and Methods: A systematic search of PubMed and National Center for Biotechnology Information (NCBI) was conducted to identify studies assessing the use of 18F-FDG PET/CT in BIA-ALCL up to and including 15 April 2024, using the following keywords “breast implant-associated anaplastic large cell lymphoma” AND “PET/CT” AND “BIA-ALCL”. Data regarding the role of PET/CT in disease detection, staging, therapeutic guidance, and post-treatment surveillance was analyzed and synthesized in a tabulated format for comparative analysis. Given study heterogeneity, findings were synthesized narratively and diagnostic performance metrics were summarized descriptively, and no formal risk-of-bias assessment was performed due to the descriptive, case-based nature of evidence. Results: A total of 28 studies met the inclusion criteria, comprising 27 individual case reports and one case series that included seven patients. Across these studies, 18F-FDG PET/CT demonstrated diagnostic utility in the evaluation of BIA-ALCL, serving primarily for initial disease staging in 27 cases and for monitoring treatment response in 16 cases. Discussion: The review’s limitations include potential search bias due to variable radiotracer terminology and the restriction to English-language studies, which may limit literature retrieval and generalizability. Conclusions: Thus, 18F-FDG PET/CT demonstrated significant value in early lesion detection, accurate staging, assisting in monitoring treatment response and detecting recurrence. Full article