The rise of online education has spurred the development of online internships, providing solutions to challenges faced by pre-service teaching trainees (PSTTs) during unique circumstances. For example, in the COVID-19 pandemic, using educational software for online internships is not bound by geographical constraints. However, there is currently a lack of research on PSTTs engaging in online internships. This study employed digital tools to facilitate the participation of PSTTs in online teaching internships and to evaluate the effectiveness of these tools. Through a combination of surveys and interviews, we investigated the post-internship attitudes of PSTTs, the factors influencing these attitudes, and their preferences for the teaching internship model. The results indicate a significant difference in PSTTs’ satisfaction based on the frequency of teacher-student interactions and the degree of collaboration. Moreover, a significant difference was observed in PSTTs’ choice of online teaching internship model based on the frequency of teacher-student interactions. The study also discusses the benefits and challenges encountered by PSTTs during their online teaching internships. The insights derived from this study provide new perspectives for online internships and teaching internships.

BACKGROUND:Technostress can harm the performance and well-being of users of information and communication, but no review has determined its global prevalence. OBJECTIVES:This systematic review (1) investigates the global prevalence of high technostress among information and communication technologies (ICT) users and (2) identifies the factors affecting the prevalence estimates. METHODS:A comprehensive three-step search was conducted across nine databases. The meta and metafor packages in R software were used to perform meta-analyses, subgroup analyses, and meta-regression analyses. The random effect model, the Hartung-Knapp-Sidik-Jonkman method, along with the Freeman-Tukey double arcsine transformation, was employed. A mixed methods appraisal tool was used to evaluate the studies' quality. Certainty of evidence was assessed. RESULTS:A total of 65 prevalence results in 61 publications involving 18,535 ICT users across 23 countries were included. The global prevalence of high technostress was 40% [95% CI: 32% to 49%]. Subgroup and meta-regression analyses revealed that country development, types of technostress, pandemic period, and the years of publication significantly influenced the prevalence estimates. CONCLUSION:Technostress requires a gold standard of definition and measure. Organizations should therefore implement preventive and protective measures to reduce the risk of developing technostress and minimize its adverse effects.

Abstract BackgroundThe issue of population aging has emerged as a critical global challenge, driving the imperative for effective self-care and scalable health management solutions for older adults. Against the backdrop of the accelerating application of generative artificial intelligence (GenAI) in health care, a systematic evaluation is necessary to investigate how multimodal GenAI can support older adults in maintaining health and managing well-being. ObjectiveThis study aimed to systematically evaluate the role, application contexts, empirical impacts, and developmental potential of diverse GenAI tools across critical geriatric health domains. MethodsA comprehensive search was executed across 11 major databases, including Web of Science, Scopus, PubMed, Medline, CINAHL, Cochrane, ACM Digital Library, IEEE Xplore, ScienceDirect, APA PsycInfo, and Google Scholar, with search transparency adhering to the PRISMA-S (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) extension. ResultsA total of 28 studies met the inclusion criteria. Of the total, 82% (n=23) of the included publications were released within the last 2 years (2024‐2025). Analysis of technology revealed that over half (n=14) of the applications were based on text-driven conversational agents, while multimodal systems, leveraging generated audio, images, and sensor data, are rapidly emerging. GenAI applications were validated to support cognitive function maintenance, mental health, and chronic condition management through personalized content generation and multimodal interaction. However, current validation is primarily limited to cognitively normal, low-risk older adult populations. Persistent technical challenges include overreliance on text-based interaction, barriers in voice recognition accuracy, and suboptimal user interface adaptability. ConclusionsPreliminary evidence suggests a promising role for GenAI in enhancing older adults’ health self-management through highly personalized and multimodal interventions, particularly in cognitive and mental health support. To realize this potential and ensure equitable access, future efforts must prioritize strengthening interdisciplinary collaboration to integrate wearable technologies and edge computing, alongside establishing robust ethical frameworks to address data privacy, algorithmic bias, and the digital divide, which will be critical to building a safe, equitable, and effective environment for active aging.

In the era of rapid advancement in digital technologies (DT), leveraging these technologies to implement health education (HE) within community spaces (CS) can have a positive impact on mitigating public health events. This study aims to conduct a systematic review of HE initiatives utilizing DT in CS, with a comprehensive analysis of target groups, types of CS, types of DT employed, and key findings. This study adheres to the PRISMA guidelines. A thorough search was conducted across three databases: Web of Science, Scopus, and PubMed. The research on the implementation of DT in community HE primarily focuses on community residents, the elderly, and community workers. The topics of HE mainly revolve around diseases, mental health, and health care. The primary implementation spaces are community centers. The main DT employed include mobile technology, digital courses and educational materials, and social media platforms. To promote the widespread application of DT in community HE in the future, it is essential to improve the usability and accessibility of these technologies, optimize educational processes, and strengthen policy support, thereby encouraging the active participation of both community workers and the public.

Generative Artificial Intelligence (AI) is rapidly changing the field of education, especially in the application of image generation technology. This study aims to systematically understand the current applications, research hotspots, and future trends of image generative AI in education through bibliometric analysis. The study is based on relevant literature from the Web of Science database between 2017 and 2024. Using CiteSpace software, it conducts a visual analysis and interpretation of trends in publication volume, prevalent keywords, keyword co-occurrence, keyword clustering, keyword bursts, timelines, time zone distribution, and country co-occurrence networks. The results show that the application of image generative AI in education has been increasing year by year. The research hotspots are mainly focused on personalized learning, art education, virtual/augmented reality, and educational assessment. There is also a trend from technical exploration to educational practice, and from single disciplines to interdisciplinary integration. Future research trends include developing image generation models more suitable for educational scenarios, designing personalized learning experiences, studying educational effect evaluation methods, and exploring ethical and social impacts. This research provides a reference for researchers in related fields and provides insights for the future application of image generative AI in education.

This study presents WordMap, an integrated text mining application developed to enhance the efficiency and usability of text analysis over a network. As unstructured text data continues to grow across domains, effective tools for segmentation and topic modeling have become increasingly essential for extracting insightful information. However, most existing solutions depend on multiple disconnected tools, and these often compromise workflow efficiency and user experience. Unlike traditional tools, WordMap combines corpus segmentation, topic modeling, and result visualization into a unified workflow for both Chinese and English languages, thereby reducing workflow fragmentation and lowering the user threshold. To assess usability and user acceptance, this research adopts the Technology Acceptance Model (TAM). WordMap employs PKUSEG and NLTK for bilingual corpus segmentation, utilizes BERTopic for dynamic topic modeling, and integrates interactive visualization to enable intuitive analysis. The PLS-SEM result shows that the perceived ease of use (PEOU) has a significant impact on both perceived usefulness (PU) and user attitude (ATT), while ATT strongly predicts behavioral intention (BI) (β = 0.674, p < 0.001). The results indicate that integrating core text mining processes into a user-centered design significantly boosts user satisfaction and adoption. By combining key processes and empirically validating user perceptions, the proposed framework facilitates the development of efficient and accessible text mining tools. It offers both theoretical and practical insights for future advancement and deployment in the field of text mining.

The rapid advancement in generative artificial intelligence (GenAI) has transformed learning, problem-solving, and creative practices in higher education, yielding a blend of both opportunities and challenges. This study investigates university students’ adoption of GenAI tools, their satisfaction with these tools, and the resulting behavioural changes by integrating the Uses and Gratifications (U&G) theory with the Technology Acceptance Model (TAM). An extended model was proposed in this study, incorporating motivational needs, technology acceptance factors, and Negative Usage Tendency (NU) as a moderating factor. A quantitative survey involving 237 university students in Macao was conducted, and the data were analysed using Structural Equation Modelling (SEM) and path analysis. Results confirmed the classical TAM pathways, indicating that Perceived Ease of Use (PEOU) influenced Perceived Usefulness (PU), which subsequently predicted Behavioural Intention (BI) and Actual Use (AU). Motivational needs were shown to have a significant impact on PEOU, while NU was found to moderate the relationship between AU and satisfaction. Collectively, these findings advanced knowledge of GenAI adoption, offering theoretical insights into university students’ thinking regarding GenAI use. Additionally, limitations and future research directions were further discussed.

BACKGROUND:Accurate diagnosis and prognosis stratification of gastric cancer (GC) are crucial for effective treatment. However, traditional histopathological image analysis relies on the subjective judgment of pathologists, which is time-consuming and prone to errors. The emergence of deep learning (DL) models provides new ways to automate and improve the analysis of GC pathology images. This systematic review aims to evaluate the current application, challenges, and future directions of DL in GC pathology image analysis. METHODS:The study follows the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Extension for Scoping Reviews (PRISMA-ScR) guidelines and searched four databases: PubMed, Scopus, Web of Science, and IEEE Xplore (as of June 19, 2025). RESULTS:The initial search identified 520 articles, and 22 studies that met the inclusion criteria were finally included. The results show that DL models have performed excellently in GC detection, histological classification, and prognosis prediction. Some models even reached an accuracy of over 95% in GC detection. Convolutional neural networks (CNN) are the most commonly used DL models. However, current studies still have limitations, such as limited dataset size, lack of external validation, and insufficient data diversity. The applicability to different types and stages of GC is also unclear. CONCLUSIONS:Future research must build larger, more diverse, and more representative datasets. These should cover a wider range of GC types and stages, and undergo rigorous clinical validation. This will help fully realize the potential of DL in GC pathology image analysis and ultimately improve clinical practice.

The intensification of global aging has made older adults’ health issues increasingly prominent. Although digital health tools play a crucial role in aging societies, older adults still face multi-dimensional potential obstacles that restrict their effective participation. Natural Language Processing (NLP) can efficiently analyze the data of older adults’ feedback in digital health to uncover hidden patterns. Therefore, this study constructs a community-based hybrid NLP analysis method to reveal the deep obstacle mechanisms affecting the digital health participation of the elderly. In this study, interview data from 35 older adults were collected through semi-structured focus group interviews in five community health activities. After pre-processing the data, a document-term matrix was constructed using Term Frequency-Inverse Document Frequency (TF-IDF) vectorization. By integrating the three methods of Latent Dirichlet Allocation (LDA) topic modeling, BERTopic topic modeling, and SBERT with K-means clustering analysis, the semantic features and topic distribution within the text data are systematically analyzed. Among them, BERTopic (K = 4) has become the most effective clustering model in this study with high word-level consistency (0.5328), reasonable topic diversity (0.5583), and good document-level consistency (0.8467). The research findings indicate that the potential obstacles affecting older adults’ digital health participation are primarily the functional and design manifestations of deeper, socio-psychological, and social environmental latent barriers. These are clustered into four aspects: psychological barriers to digital technology use, clarity and usability of health information, a mismatch between technology provision and user needs, and user interface (UI) design and accessibility challenges. Older adults’ relatively low digital literacy leads to fear and anxiety toward technology. In obtaining effective information, they face the issue of information overload and have doubts about the authenticity and reliability of the information. The lack of positive technology learning experiences and external support makes older adults prone to a sense of low self-efficacy. The uneven distribution of community resources and the limited technical support capabilities of family members further exacerbate the obstacles to older adults’ digital health participation. This study proposes a new method for in-depth exploration of the potential obstacles to older adults’ digital health participation from a semantic level by combining multiple clustering analysis methods. It provides valuable references and technical guidance for subsequent related research.

Flight cancellation prediction accuracy remains essential for airlines because it allows for automatic risk reduction of financial losses and passenger satisfaction decline. Heavy aviation big data presents challenges to traditional prediction methods which makes their practical use difficult. The proposed research brings forth an innovative approach utilizing distributed ensemble learning for conducting flight cancellation predictions at scale. The Artificial Bee Colony (ABC) algorithm operates within our method to determine the most essential predictors from an extensive dataset through optimal feature selection. The MapReduce framework enables distributed K-Nearest Neighbor (DKNN) model implementation to process features selected by the subsequent stage. The distribution of KNN models within this architecture allows the processing of extensive datasets effectively and delivers better accuracy through a collective model voting system. Our system performs computations on flight data collected from three New York City airports (JFK, LGA, and EWR) with a minimum computational advantage exceeding 25% above non-distributed KNN models. The ensemble strategy enhances prediction accuracy by 3.42% to obtain an average accuracy level of 95.79% which represents a 2.2% improvement above previous methods. Our distributed ensemble methodology proves its effectiveness for predicting flight cancellations accurately in big data environments through the presented experimental results.

Learning through short videos (SVs) is an effective medium for international students (ISs). This active learning method allows students, particularly those from abroad, to quickly and conveniently gain knowledge about life and study through short video content. Using Web of Science (WOS) data, CiteSpace analyzes the spatial patterns, trends, and hotspots within research studies on ISs utilizing SVs for learning. Results: 1) ISs and SVs initially increased in the first half of the study period, reaching a peak before experiencing a decline over the next five years. This trend indicates a predominant focus on the ISs SVs field from the perspectives of education and sociology. 2) This research has formed a complex international collaboration network, particularly among the United States, Spain, and China. 3) Educational studies examining the adaptability of ISs and issues such as culture shock, particularly those concerning SVs based on digital technologies and tools, have emerged as a primary focus area. 4) With the relaxation of study abroad policies in various countries and the further development of short video platforms, research on ISs and SVs is likely to regain momentum. Conclusion: This study provides a comprehensive and objective research analysis of ISs and SVs. This highlights the role and impact of SVs as a form of digital educational technology on the learning experiences of ISs.

Treatment-induced ovarian function loss is a significant concern for many young patients with breast cancer. Accurately predicting this risk is crucial for counselling young patients and informing their fertility-related decision-making. However, current risk prediction models for treatment-related ovarian function loss have limitations. To provide a broader representation of patient cohorts and improve feature selection, we combined retrospective data from six datasets within the FoRECAsT (Infertility after Cancer Predictor) databank, including 2679 pre-menopausal women diagnosed with breast cancer. This combined dataset presented notable missingness, prompting us to employ cross imputation using the k-nearest neighbours (KNN) machine learning (ML) algorithm. Employing Lasso regression, we developed an ML model to forecast the risk of treatment-related amenorrhea as a surrogate marker of ovarian function loss at 12 months after starting chemotherapy. Our model identified 20 variables significantly associated with risk of developing amenorrhea. Internal validation resulted in an area under the receiver operating characteristic curve (AUC) of 0.820 (95% CI: 0.817-0.823), while external validation with another dataset demonstrated an AUC of 0.743 (95% CI: 0.666-0.818). A cutoff of 0.20 was chosen to achieve higher sensitivity in validation, as false negatives-patients incorrectly classified as likely to regain menses-could miss timely opportunities for fertility preservation if desired. At this threshold, internal validation yielded sensitivity and precision rates of 91.3% and 61.7%, respectively, while external validation showed 92.9% and 60.0%. Leveraging ML methodologies, we not only devised a model for personalised risk prediction of amenorrhea, demonstrating substantial enhancements over existing models but also showcased a robust framework for maximally harnessing available data sources.

Since the release of OpenAI's ChatGPT in 2022, AI activity has reached a fever pitch. Calls for effective ethical responses to the pressurised AI environment have in turn abounded. Posthumanism, which seeks to build ethical futures by de-centring the ‘human’, is an obvious candidate to act as a lynchpin of theoretical intervention. In their responses, posthumanist scholars appear to have embraced AI’s potential to destabilise Humanist philosophical ideas. We critically interrogate this initial enthusiasm. Conceptually distinguishing ‘post-dualist self-development’ (PDSD) from ‘technical self-development’ (TSD), we show how AI prompts an urgent need to advance posthumanist engagement with how technical development unsupervised by humans is ontologically discrete from other forms of material agency. We argue that specific engagement with TSD as distinct from PDSD is a key to avoid ignoring or underestimating Humanist and anthropocentric aspects of current AI innovation, and the influence of anthropomorphism. Without a theoretical reckoning with these tensions, posthumanism in the AI-era runs the risk of potentially promoting technologies that reinvigorate Humanist and anthropocentric expansion. To conclude, we show how a posthumanist ethics of generative AI that pays requisite attention to both TSD and PDSD may enable more anticipatory and nuanced assessments of the risks and benefits of discrete AI technologies to inform public discourse, appropriate social, institutional, policy and governance responses, and direct AI research and development priorities.

BackgroundKidney stones, a prevalent urinary disease, pose significant health risks. Factors like insufficient water intake or a high-protein diet increase an individual’s susceptibility to the disease. Social media platforms can be a valuable avenue for users to share their experiences in managing these risk factors. Analyzing such patient-reported information can provide crucial insights into risk factors, potentially leading to improved quality of life for other patients. ObjectiveThis study aims to develop a model KSrisk-GPT, based on a large language model (LLM) to identify potential kidney stone risk factors from web-based user experiences. MethodsThis study collected data on the topic of kidney stones on Zhihu in the past 5 years and obtained 11,819 user comments. Experts organized the most common risk factors for kidney stones into six categories. Then, we use the least-to-most prompting in the chain-of-thought prompting to enable GPT-4.0 to think like an expert and ask GPT to identify risk factors from the comments. Metrics, including accuracy, precision, recall, and F1-score, were used to evaluate the performance of such a model. ResultsOur proposed method outperforms other models in identifying comments containing risk factors with 95.9% accuracy and F1-score, with a precision of 95.6% and a recall of 96.2%. Out of the 863 comments identified with risk factors, our analysis showed the most mentioned risk factors for kidney stones in Zhihu user discussions, mainly including dietary habits (high protein, high calcium intake), insufficient water intake, genetic factors, and lifestyle. In addition, new potential risk factors were discovered with GPT, such as excessive use of supplements like vitamin C and calcium, laxatives, and hyperparathyroidism. ConclusionsComments from social media users offer a new data source for disease prevention and understanding patient journeys. Our method not only sheds light on using LLMs to efficiently summarize risk factors from social media data but also on LLMs’ potential to identify new potential factors from the patient’s perspective.

Virtual reality (VR) technology is revolutionizing the landscape of digital health education by providing learners with increasingly immersive and interactive experiences. This study aims to understand the current usage, main research topics, and future directions of VR in digital health education through a bibliometric analysis. Using data from the Web of Science Core Collection from 2008–2024, alongside CiteSpace software, this study examined trends in publication volume, prevalent keywords, keyword co-occurrence, clustering, keyword bursts, timelines, time-zone distributions, and author networks. Research on VR in digital health education has increased over the years. The main topics are medical simulation training, health education, and rehabilitation therapy. Additionally, the study demonstrated a transition from mere exploration of the technology to its widespread application across diverse fields. Future trends may include creating more immersive and interactive VR education platforms, designing personalized learning experiences, studying how to measure the effects of VR education, and discussing ethical concerns. This research provides a reference for others in the field and provides insights into future VR applications in digital health education.

Large Language Models (LLMs) present paradigm-shifting potential in mental health, offering opportunities to streamline clinical workflows and broaden investigative frameworks. This study reviews recent literature at the intersection of LLM and mental health, categorizes and analyzes different research types, identifies key limitations in current studies, and proposes forward-looking directions for future research. We conducted a systematic review of LLMs in mental health research (January 2023-March 2025) by searching seven academic databases using specific keyword combinations related to mental health and LLMs. The selection criteria prioritized peer-reviewed empirical studies, high-quality preprints, and expert analyses while excluding brief commentaries and narrative reviews without systematic methodologies. Our analysis categorized the research into six main areas: machine psychology, diagnosis, treatment, clinical assistants, research and education, as well as applications of multimodal techniques. We found that LLMs, despite their considerable utility, face dual categories of challenges: general technical limitations (including content hallucination and inadequate privacy safeguards),and domain-specific constraints (including suboptimal diagnostic accuracy and limited professional depth in specialized contexts) In response to these challenges, we delineate solutions such as deepening clinical research, applying new technologies, establishing practice guidelines and enhancing interpretability, and propose short- and long-term development strategies. This review provides a systematic framework for understanding the current status and prospects of the application of LLMs in mental health, and promotes the integration of technological innovation and humanistic care.

Integrating Elasticsearch as the retriever in a Retrieval-Augmented Generation (RAG) framework is essential for improving chatbot responses’ factual accuracy and contextual relevance. Traditional Large Language Models (LLMs) often fall short of providing up-to-date information, especially in dynamic domains like training. To meet the demands of smart training, Elasticsearch is integrated into the RAG framework as the retrieval module to enable efficient access to up-to-date and context-specific information. System-level experiments and a user-centered survey have been conducted to evaluate the effectiveness of system. Technical performance is assessed through comparisons with traditional retrieval methods, while user acceptance is evaluated by using the Technology Acceptance Model (TAM). System-level experiments show that integrating Elasticsearch into the RAG framework substantially advances chatbot performance, with the F1-score increasing from 0.47 to 0.85 and response time decreasing from 1.5 to 0.8 seconds. User satisfaction improves from 31% to 82%. The user-centered survey based on the TAM confirms the system’s effectiveness, with all four constructs demonstrating acceptable reliability (Cronbach’s α > 0.65) and composite reliability (> 0.79). Multiple regression analysis reveals that Attitude Toward Use (ATT; β = 0.280, p < .001) and Perceived Usefulness (PU; β = 0.279, p < .001) significantly predict users’ Behavioral Intention (BI) to use the system. These findings indicate that integrating Elasticsearch into the RAG framework advances retrieval accuracy and efficiency. Furthermore, it enhances user acceptance in bilingual smart training chatbot applications, improves the learning experience, and offers a practical approach to advancing smart training.

Exploring students' cognitive abilities has long been an important topic in education. This study employs data-driven artificial intelligence (AI) models supported by explainability algorithms and PSM causal inference to investigate the factors influencing students' cognitive abilities, and it delved into the differences that arise when using various explainability AI algorithms to analyze educational data mining models. In this paper, five AI models were used to model educational data. Subsequently, four interpretable algorithms, including feature importance, Morris Sensitivity, SHAP, and LIME, were used to globally interpret the results, and PSM causal tests were performed on the factors that affect students' cognitive abilities. The results reveal that self-perception and parental expectations have a certain impact on students' cognitive abilities, as indicated by all algorithms. Our work also uncovers that different explainability algorithms exhibit varying preferences and inclinations when interpreting the model, as evidenced by discrepancies in the top ten features highlighted by each algorithm. Morris Sensitivity presents a more balanced perspective, while SHAP and feature importance reflect the diversity of interpretable algorithms, and LIME shows a unique perspective. This detailed observation highlights the practical contribution of interpretable AI algorithms in the field of educational data mining, paving the way for more refined applications and deeper insights in future research.

Educational Data Mining (EDM) supports early detection of learning difficulties by predicting student performance. However, machine learning models often operate as black boxes. Explainable Artificial Intelligence (XAI) helps to explain why black-box models produce specific predictions. This paper systematically reviews the past five years of research on XAI applications for interpreting student performance prediction in Computer Science and STEM education. We found that behavioral and academic performance data were the most commonly used features, with the main prediction goals focused on course failure risk or grades. This study also examined the application areas of XAI, revealing that the most common uses were global feature importance analysis, individual prediction explanations, and supporting interventions and decision-making. Moreover, we found that SHapley Additive exPlanations (SHAP) were the most frequently utilized XAI technique, predominantly applied at the global level, with limited use at the individual level. Furthermore, a research gap was identified in utilizing XAI to support course improvements, customize visualizations, and generate personalized recommendations. Addressing this gap could enable educators to provide personalized, data-driven guidance to better support individual students.

Sentiment analysis using Large Language Models (LLMs) has gained significant attention in recent research due to its outstanding performance and ability to understand complex texts. However, popular LLMs, such as ChatGPT, are typically closed-source and come with substantial API costs, posing challenges for resource-limited scenarios and raising concerns about privacy. To address this, our study evaluates the feasibility of using small LLMs (sLLMs) as alternatives to GPT for aspect-based sentiment analysis in Chinese healthcare reviews.We compared several Chinese sLLMs of varying sizes with GPT-3.5, using GPT-4o’s results as the benchmark, and assessed their classification accuracy by computing F1 scores for each individual aspect as well as an overall F1 score. Additionally, we examined sLLMs’ instruction-following capabilities, VRAM requirements, generation times, and the impact of temperature settings on the performance of top-performing sLLMs. The results demonstrate that several sLLMs can effectively follow instructions and even surpass GPT-3.5 in accuracy. For instance, InternLM2.5 achieved an F1 score of 0.85 with zero-shot prompting, while the smaller Qwen2.5-3B model performed well despite its minimal size. Prompt strategies significantly influenced smaller and older models like Qwen2.5-1.5B and ChatGLM3.5 but had limited impact on newer models. Temperature settings showed minimal effect, while older models generated responses faster, and newer models offered higher accuracy. This study underscores the potential of sLLMs as resource-efficient, privacy-preserving alternatives to closed-source LLMs in specialized domains. Our work demonstrates versatility, with potential applications across domains such as finance and education, and tasks like sentiment analysis, credit risk assessment, and learning behavior analysis, offering valuable insights for real-world use cases.