[Purpose/Significance] As artificial intelligence (AI) systems are being widely deployed across diverse domains such as education, healthcare, and public governance, the absence of standardized metadata specifications has led to fragmented descriptions, inconsistent documentation, and difficulties in model evaluation and reuse. This study aims to address the pressing issues of opacity, lack of interpretability, and poor traceability in current AI models, which have increasingly become obstacles to the development of transparent and responsible AI. To overcome these challenges, this study proposes the establishment of a unified metadata specification for AI models to enhance their discoverability, transparency, interoperability, and reusability, thereby advancing the development of trustworthy AI and facilitating effective model governance. [Method/Process] Grounded in metadata quality assessment theory and lifecycle theory, the study adopted a combination of research methods, including literature review, comparative analysis of existing specifications, and questionnaire surveys.We first conducted a systematic examination of domestic and international practices related to AI model metadata specifications to identify representative standards, frameworks, and implementation approaches. Through comparative analysis, the study investigated the structure, element organization, and semantic relationships of different specifications, highlighting their similarities, differences, and areas for improvement. Meanwhile, a targeted questionnaire survey was administered to researchers, developers, and practitioners to explore user awareness, perceptions, practical experiences, and specific needs regarding metadata specification and interoperability. Based on these findings, the study ultimately proposed a lifecycle-oriented framework for metadata specification construction, ensuring that it aligns with the key stages of AI model development, deployment, evaluation, and governance. [Results/Conclusions] The findings reveal that, although users generally recognize the importance of metadata specifications for AI models, they are unaware of of the existing specifications. The current AI model metadata specifications have significant shortcomings in terms of element naming, structural organization, and descriptive granularity. These shortcomings hinder the effective sharing and reuse of model information. In response, the study proposed a comprehensive metadata framework encompassing key entities such as models, datasets, algorithms, technical features, performance evaluations, risks and ethics, legal information, and related resources, as well as the semantic relationships among these entities. The research concluded that establishing a unified metadata specification for AI models not only contributes to effective information management and cross-platform interoperability, but also serves as a critical infrastructure that links technology, ethics, and governance. As the metadata specification system matures and gains wider industry adoption, AI models will become increasingly controllable and trustworthy. This will promote a more regulated, collaborative, sustainable and integrated AI ecosystem.

[Purpose/Significance] Against the backdrop where artificial intelligence generated content (AIGC) is reshaping the paradigm of knowledge production, exploring how to integrate the context-aware capabilities of large models into the knowledge service framework, and on this basis, providing a new interpretation and service expansion of the Five Laws of Library Science, is of great significance for the construction of the AI + knowledge service system. [Method/Process] Starting from the perspective of contextual integration, and with Wilson's Information Behavior Theory and the SECI Model as the theoretical foundation, this study constructs a three-dimensional integrated coupling framework of "demand-knowledge-context". By aligning knowledge context with user context, it dynamically perceives users' knowledge needs and provides proactive responsive services. Based on this framework, it further conducts a contextualized interpretation of the Five Laws of Library Science. [Results/Conclusions] With the integration of all contextual elements as the link, the "demand-knowledge-context" three-dimensional framework forms an effective mechanism for matching user needs with knowledge resources. It also achieves continuous reinforcement learning through information feedback to acquire the ability of self-evolution, aiming to continuously adapt to and better meet users' knowledge needs in complex and changing contexts. Based on this framework, the study conducts contextualized interpretation and theoretical expansion of the Five Laws, endowing library services with new connotations of vitality and intelligence. Furthermore, it proposes a context-driven ecological evolution path for knowledge services: the internal context focuses on the recombination of organizational genes, while the external context emphasizes the dissolution of ecological boundaries, exploring how classic library theories can achieve innovative development with the support of new technologies. As AIGC technology continues to develop, further in-depth research into contextual elements should be conducted, particularly into implicit contextual elements such as users' emotions and psychology. Efforts should be devoted to strengthening interdisciplinary collaboration, incorporate theories and methods from disciplines such as psychology and sociology into library science research, and continuously optimize the knowledge service system of smart libraries. This will make the system more adaptable to complex and changing contexts, provide users with higher-quality, more efficient, and personalized knowledge services, and help library science achieve significant development in the AIGC era. This paper proposes a three-dimensional "demand-knowledge-context" framework that aims to accurately match user needs with knowledge resources by integrating knowledge and user context in depth. Based on this, the "context-driven ecological evolution approach for knowledge services" is put forward as an exploratory implementation approach. However, subsequent research is required to implement, verify and conduct an in-depth empirical analysis of this approach.

[Purpose/Signficance] In the context of the increasingly widespread adoption of digital communication, agriculture-related emergencies often trigger complex and ever-changing public opinion online due to their high level of specialization and the significant cognitive barriers they pose to the general public. Emotional factors play a pivotal role in the evolution and governance of online public opinion. However, current research into how public opinion is guided in relation to agricultural emergencies still fails to systematically address emotional factors. [Method/Process] Therefore, the study constructed an analytical framework for emotional guidance in agricultural-related public opinion, integrating information subjects, information content, and the information environment. The framework was based on three complementary theories: information ecology theory, social amplification of risk theory, and negativity bias theory. It explored the correlations and combined effects of emotional factors with individual audiences, media, and the information environment. A total of 31 online public opinion cases involving agriculture, rural areas, and farmers were selected from the "Public Opinion Daily Reports" published by the People's Daily Online Public Opinion Data Center, covering the period from January 2021 to June 2025. The Weibo platform was chosen for this study, and data were collected by searching for case names and related topics on Weibo to capture raw data for conditional and outcome variables. Sentiment analysis was introduced to identify and quantify emotional characteristics in public opinion, and fuzzy-set qualitative comparative analysis (fsQCA) was employed to investigate how various factors collectively influence the guidance of online public opinion in public emergencies. The aim is to reveal the emotional guidance mechanisms and the logic behind effect formation in online public opinion regarding agricultural emergencies. [Results/Conclusions] The study found that public opinion in agriculture exhibits typical characteristics of equifinal multiple causation, whereby various combinations of factors can produce similar guiding effects. In contexts of high emotional polarisation, the pathways may rely on traffic restriction and emotional substitution regulation. In contexts of low emotional polarization, they may rely on the construction of emotional framing by authoritative media and opinion leaders. In different contexts, information clarity and netizens' emotional involvement can form a substitution relationship with the degree to which the platform intervenes in emotional regulation. This necessitates dynamic adjustments to guidance strategies based on specific situations. Based on this, the governance of agriculture-related public opinion online should shift towards a systematic emotional governance framework that leverages affective computing to expand the range of channels and strengthen the basis of public opinion. Efforts should also be devoted to strengthening dynamic response mechanisms based on real-time emotional monitoring. The aim should be to construct a sentiment guidance system for public opinion featuring dynamic allocation, multi-party collaboration, and precise reach.

[Purpose/Significance] In recent years, the rapid rise of large language model technology has shown significant advantages in understanding semantic context and capturing multidimensional sentiment tendencies. This study explores an aspect-level sentiment analysis method for science and technology policy comments based on large language models, aiming to uncover latent knowledge within these texts and provide data support for evaluating the effectiveness and subsequent optimization of policies. [Method/Process] Taking the electric vehicle industry as an example, a burgeoning sector vital to achieving the "dual carbon" goals and promoting green low-carbon development, this study proposed a policy satisfaction evaluation model. The model uses large language models for fine-grained aspect-level sentiment analysis of policy comment texts. The process includes the following steps: 1) Data collection and preprocessing: Comments related to electric vehicle policies were collected from the "Interactive Topics" section of the "Autohome" website using Python. Deep learning techniques were applied to set rules for the comment texts and automatically add punctuation marks to Chinese texts for data pre-processing. 2) Aspect word extraction: The steps include text tokenization, determining a candidate aspect word set, expanding the aspect word set, and clustering aspect words. A total of 3 405 aspect words were extracted from 35 000 comments, forming six clusters: infrastructure construction, vehicle performance configuration, national policies, technological development, automotive safety, and automotive sales market. Aspect-level sentences were extracted using aspect words and punctuation information, with a subset of sentences manually labeled to build training and validation corpora, resulting in 14 911 aspect-level sentences. 3) Sentiment tendency recognition model training: A prompt template for aspect-level sentiment classification tasks was designed, and the LoRA method was used to fine-tune the large language model with the manually labeled training set. The model's performance was evaluated using a validation set, resulting in the classification of comments on electric vehicle policies into positive, neutral, and negative sentiments. 4) Comparative experiment: The fine-tuned large model was compared with the mainstream sentiment classification model, BERT, to assess the performance of different models in aspect-level sentiment classification tasks. [Results/Conclusions] The results show that compared to the BERT model, the proposed method outperformed other methods in multiple metrics, including accuracy, recall, and F1 score, with improvements of 11.49%, 12.43% and 11.43%, respectively. Overall, public attention is higher towards vehicle performance configuration and automotive sales market, while infrastructure construction receives the lowest attention. The overall public satisfaction with electric vehicles is relatively low, with negative comments outweighing positive comments across all aspects, consistent with the "negative bias" theory in social psychology. Satisfaction issues are particularly prominent in the areas of automotive safety and infrastructure construction. Finally, policy recommendations have been proposed to optimize electric vehicle subsidy policies, strengthen policy promotion, improve infrastructure construction, and enhance after-sales service support systems.

[Purpose/Significance] The rapid advancement of digital infrastructure has precipitated a fundamental transformation in scholarly communication, characterized by an increasing reliance on online platforms. Preprint exchange, as a cornerstone of open science, offers researchers opportunities for immediate dissemination and collaborative engagement. However, the absence of rigorous peer review raises persistent concerns regarding research ethics, data integrity, and the reliability of scholarly outputs, which can undermine public confidence in preprint platforms. Addressing these challenges is essential not only for maintaining the integrity of academic discourse but also for fostering a transparent and trustworthy open science ecosystem. This research contributes to the existing scholarship by systematically examining the trust framework of preprint platforms, positioning itself at the intersection of library and information science and scholarly communication studies. In contrast to previous investigations that have focused predominantly on dissemination efficiency or platform functionality, this study emphasizes the structural dimensions of trustworthiness. It presents an innovative analytical framework that strengthens the theoretical foundations of academic information trust and provides practical strategies for enhancing the governance and legitimacy of preprint platforms. [Method/Process] To ensure both theoretical rigor and empirical depth, first, a comprehensive literature review was conducted to identify potential trust-related vulnerabilities in preprint platforms and to systematically delineate their credibility challenges. This review identified five critical factors influencing the credibility of preprint platforms: academic conflicts of interest, platform reliability, heterogeneous manuscript quality, information overload, and insufficient academic recognition. Drawing upon the DeLone & McLean (D&M) Information Systems Success Model and aligning with the ISO 16363 standard for trustworthy digital repositories, the study analyzed the structural components of trustworthiness through the dimensions of system quality, information quality, and service quality. Subsequently, in-depth case studies of prominent platforms, including arXiv and ChinaXiv, were undertaken to examine their governance architectures, operational methodologies, and practical implementations. This process culminated in evidence-based recommendations for enhancing platform trustworthiness. This integrated methodological framework not only synthesizes theoretical insights with empirical evidence but also ensures the scientific rigor, reliability, and practical applicability of the proposed trust model. [Results/Conclusions] Based on these findings, a three-dimensional trust framework was developed, encompassing system trustworthiness, information trustworthiness, and service trustworthiness. This framework transcends traditional quality control paradigms and offers novel perspectives for the standardized development of preprint platforms. The research further articulates pathways for establishing trustworthiness across three levels: 1) system trustworthiness, adhering to FAIR principles and implementing long-term preservation strategies to provide a stable institutional foundation; 2) information trustworthiness, establishing a comprehensive quality governance continuum that incorporates "pre-screening, dynamic identification, and post-peer review" mechanisms; and 3) service trustworthiness, delivering professional preprint services through collaborative governance models and journal coordination frameworks.While this framework provides a comprehensive analytical perspective, certain limitations should be acknowledged. This study's primary reliance on qualitative methods necessitates broader empirical validation. Furthermore, its focus was on platform functionalities rather than user perceptions. Consequently, future research can adopt a mixed-methods approach, incorporate user perception theories, and establish quantitative metrics for evaluating platform trustworthiness.

[Purpose/Significance] Science and technology have emerged as pivotal domains of competition between China and the United States. This article provides a quantitative analysis of US technology think tanks reports on the electronic information research and industry, with a focus on the evolution of themes and topics over the past decade. This analysis not only reflects their technological priorities but also maps their analytical focus on China, providing decision-making support for China's think tanks development and strategic response. [Method/Process] Based on the "2020 Global Go to Think Tank Index Report" released by the Think Tanks and Civil Societies Program (TTCSP) at the University of Pennsylvania, considering factors such as think tank authority, research topic relevance, and research continuity, we collected a total of 1 360 reports on the electronic information research and industry published between 2015 and 2024 by 8 leading US technology think tanks. Topic analysis was conducted with BERTopic, a topic modeling tool based on Transformer embeddings. The methodology involved several key steps. First, text cleaning was performed using NLTK tools; then, the all-MiniLM-L6-v2 model was employed to generate high-dimensional document embedding vectors. Subsequently, dimensionality reduction was achieved through the UMAP algorithm, followed by density clustering using the HDBSCAN algorithm. Finally, topic words were extracted based on the c-TF-IDF algorithm. [Results/Conclusions] The research identified 31 distinct research themes, of which 6 were directly related to China, specifically: global semiconductor industry competition, Sino-US digital policies and cloud computing competition, 5G network and technology competition, Chinese AI investment, Sino-US science and innovation policies, and Sino-US military technology competition. These 31 research themes were hierarchically clustered using HDBSCAN and could be categorized into 11 major research directions. The US technology think tanks persistently focused on 11 major research directions, which were largely concentrated on key areas of electronic information research and industry, such as semiconductors and microelectronics, artificial intelligence, wireless communication, quantum information technology, network security, and big data. The evolutionary trends across these research directions were generally consistent, with military technology and network security receiving the highest level of attention. The attention attached to China has undergone a significant strategic shift over the years, with drastic increase in semiconductor export control, AI technology and Sino-US digital competition. Based on the identified key themes and topic words, it is highly recommended to establish an evolutionary mapping of China-related topics and to develop a dynamic monitoring and early warning mechanism for technology issues concerning China. Future research could incorporate larger-scale corpus resources and more advanced large language models to continuously optimize topic modeling effectiveness.

[Purpose/Significance] Artificial intelligence is reshaping the landscape of higher education, driving the transformation of university library resource development from a "resource supply-oriented" model to an "efficiency-driven" one. Traditional evaluation systems, constrained by single-dimensional indicators and manual data collection, fail to meet the demands of intelligent transformation. This study takes the Library of Beijing Institute of Technology (BIT), a research university library, as a case study to construct a three-dimensional efficiency evaluation system for resource development. Unlike previous studies that focus on theoretical model construction, this research integrates the practical experience of the BIT Library since 2021, combining team building, technological empowerment and institutional design to form an implementable evaluation system, which provides replicable references for the intelligent development of resource construction in domestic university libraries. [Method/Process] This study adopts a mixed research method combining case study and quantitative-qualitative analysis, with theoretical foundations in library science theories such as resource life cycle management and interdisciplinary theories including artificial intelligence application. The empirical data are derived from the operational data of the BIT Library from 2021 to 2025 (such as procurement records and user behavior data). The construction of the evaluation system consists of five links: 1) We establish a hierarchical training system covering all librarians, offering expert lectures and implementing the pairing model of "data analyst + subject librarian"; 2) We use the analytic hierarchy process (AHP) to determine indicator weights, adding the "interdisciplinary adaptability" indicator for emerging fields, and building a full-process evaluation model; 3) We construct an AI-empowered platform integrating multi-source data, which significantly shortens the duration of manual data processing; 4) We carry out in-depth research in collaboration with colleges and research teams, and conduct benchmarking analysis with top university libraries; 5) We establish a scientific decision-making mechanism linking evaluation data with the University Library Committee and various colleges of the university. [Results/Conclusions] The application of this system in the BIT Library has achieved remarkable results: the accuracy of library resource guarantee has been significantly improved, the efficiency of resource utilization has risen substantially, the capacity for scientific research support has been notably enhanced, and the service satisfaction has undergone a leapfrog improvement. However, the evaluation system still has limitations such as data privacy risks, insufficient AI literacy of some librarians, and lack of inter-library collaboration. Suggestions for targeted action include adopting federated learning technology to protect data privacy, carrying out hierarchical AI training, and establishing a regional evaluation alliance. Future research will explore the specific application of generative artificial intelligence in the evaluation system and establish a dynamic adjustment mechanism for indicators adapted to technological and disciplinary development.