[Purpose/Significance] To explore the complex causal relationship between think tank brands and their influence is of great significance for the brand construction, international discourse power establishment, and influence enhancement of Chinese think tanks. [Method/Process] Based on the brand identity system, this study built the brand elements of think tanks. Analytic hierarchy process(AHP) and the expert scoring method were used to determine the weights of think tank brand elements. The data obtained from the think tanks' official websites, thematic databases, and other channels were normalized, weighted, and aggregated into the conditional variables of brand-driven think tank influence enhancement. Fuzzy-set qualitative comparative analysis(fsQCA) was used to explore the complex causal relationships among these conditional variables and the influence enhancement of think tanks. [Result/Conclusion] It is found that the conditional variables of brand-driven think tank influence enhancement are think tank symbol, think tank service, think tank organization, and think tank culture. The enhancement paths of think tanks' influence are diversified. There are two paths of foreign well-known think tanks: one jointly driven by think tank symbol and think tank culture, and another jointly driven by think tank service and think tank organization. There are three paths in well-known domestic think tanks: one driven by think tank symbol, one jointly driven by think tank symbol, think tank organization and think tank culture, and one driven by think tank culture. These five paths provide a reference for Chinese think tanks to build brands, strengthen global discourse power, and expand international influence.

[Purpose/Significance] Current methods for identifying transferable university patents face challenges in accurately constructing inventor features, particularly for patents with multiple inventors. This study proposes a research team feature-based approach to improve the identification accuracy. [Method/Process] First, an optimized label propagation algorithm is used to identify university research teams. Then, based on the identified teams, a method for constructing research team features centered on core members is proposed. Finally, these team features is integrated into a machine learning model to identify transferable patents, and the SHAP model along with decision rules are employed to interpret the results. [Result/Conclusion] The effectiveness of the proposed method is validated through experiments on patent transfer data from Nanjing Tech University and Stanford University. This approach provides technical support for the transferable university patents.

[Purpose/Significance] Scientific experiments are the core process of scientific research and innovation. To identify and model their key elements and relationships will help reveal the inherent laws of research innovation activities. This study proposes an entity-relationship conceptual model of innovation elements for scientific experiment processes, which is conducive to structuring the expression of scientific research knowledge and identifying the innovation paths. [Method/Process] First, this study analyzed the formation mechanism of scientific-data-driven technological innovation and deconstructed the types and implementation processes of scientific experiments. It clarified the scientific data elements and technological innovation elements involved in scientific experiments, and constructed a conceptual model comprising 5 types of element entities and 8 types of entity relationships by using a quintuple structure to express. Then, a specific entities and semantic relationship system was constructed by taking rice breeding as an example, combined with domain database rules and expert knowledge. Therefore, a domain innovation path model based on scientific experimental events was formed to demonstrate the practical application of the model. [Result/Conclusion] This model can characterize the key elements and relationships in the scientific research process in a fine-grained manner. The innovation path based on this model can reveal the innovation process and laws driven by single-type as well as multi-type innovation factors, laying a foundation for the discovery of innovation opportunities from scientific research processes.

[Purpose/Significance] The rapid evolution of next-generation artificial intelligence technologies has triggered a series of complex knowledge interaction challenges. It becomes a critical issue in advancing new quality productive forces through technological innovation. Investigating novel human–AI relationships and constructing theoretical models of their knowledge assimilation is of great theoretical and practical significance for promoting the development of knowledge management in interaction contexts. [Method/Process] This study adopted the lens of human-AI interaction to examine knowledge assimilation under emerging human-AI relationships. It identified key dimensions of the assimilation process, including knowledge agents, modes of knowledge representation, knowledge attributes, interaction patterns, and core features. On this basis, a theoretical model of dual-agent knowledge assimilation between human and artificial intelligence entities was proposed. [Result/Conclusion] Human-AI relationships are embedded in both exchange-based and co-ownership-oriented interpersonal dimensions, characterized by a duality of independence and interdependence. The evolving dynamics of human-AI interaction, together with heterogeneity in knowledge representation, give rise to new knowledge production mechanisms that integrate explicit knowledge, tacit knowledge, latent knowledge, and domains of ignorance. Four distinct modes of knowledge assimilation emerge—search-based, supply-based, embedded, and symbiotic—each associated with different knowledge demands and exhibiting dynamic mode-shifting effects. Ultimately, dual-agent knowledge assimilation is found to be a dynamic, non-linear evolutionary process. These findings offer valuable insights for fostering intelligent, collaborative, and creative interactions between individuals, organizations, and advanced AI systems, thereby advancing Human–AI innovation and knowledge co-creation.

[Purpose/Significance] This study explores the application of fine-tuned large language models(LLMs) in scientific entity recognition from academic literature, aiming to improve the recognition performance of specific types of scientific entities. [Method/Process] First, the original evaluation dataset was preprocessed by sentence segmentation and entity filtering. Specific prompt templates were designed to convert the preprocessed data into a dataset compatible with the fine-tuning format of LLMs. Subsequently, the Low-Rank Adaptation(LoRA) strategy was employed to conduct supervised fine-tuning on four LLMs from the DeepSeek and Qwen series. These models were compared with state-of-the-art(SOTA) deep learning models and prompt learning methods based on the DeepSeek-R1-671B model. Performance was comprehensively evaluated using precision, recall, and F1-score, combined with efficiency metrics such as training runtime and floating-point operations. Finally, an in-depth analysis of the models' misidentification types was conducted. [Result/Conclusion] Experimental results demonstrate that the fine-tuned LLMs perform significantly in scientific entity recognition tasks, achieving an F1 score of 81.20%, and effectively improving the accuracy of scientific entity recognition.

[Purpose/Significance] This study aims to clarify the concept and identification methods of innovation in academic papers. It explores the application paths to provide references for constructing sustainable academic innovation ecosystem. [Method/Process] Based on the concept of innovation in academic papers, it summarizes its main forms of innovation, conceptualizes its transformation process, and systematically reviews the existing identification methods. Furthermore, it discusses the application path for the innovation. [Result/Conclusion] It identifies two major methodological systems based on text mining and scientific knowledge network. It designs a systematic application path: “content insight→significance cognition→frontier prediction→ex-post evaluation”. Based on the analysis, it proposes the following constructive considerations: building a multimodal scientific knowledge base from multiple sources, improving early warning mechanisms for innovation, deepening the multidimensional discovery system, and reforming scientific evaluation mechanisms.

[Purpose/Significance] In the era of data elements, the innovation in public data development and utilization has promoted the development of new quality productive forces. Exploring its innovation mechanism is conducive to unlocking the value of public data. [Method/Process] Focusing on the innovation mechanism of public data development and utilization, and based on grounded theory, interviews and data coding analysis were conducted with 18 innovation participants. Consequently, an innovation mechanism model for public data development and utilization was constructed, covering 4 core categories, 12 main categories, 46 basic categories, and 170 initial concepts. [Result/Conclusion] The research finds that the innovation in public data development and utilization is a dynamic process of “trigger-evolution-regulation-guarantee”, which includes four stages: idea generation, refinement and verification, implementation support, and implementation diffusion. Its influencing factors include innovation-triggering factors, innovation-regulating factors, and innovation-guaranteeing factors. The research systematically analyzes the process logic and operational mechanism of each element in the innovation in public data development and utilization, providing a theoretical framework and practical guidance for deepening the development and utilization of public data.

[Purpose/Significance] This study explores the cognitive characteristics of users in AIGC multi-modal information search, aiming to provide references for optimizing AIGC multi-modal search tools and enhancing users’ search capabilities. [Method/Process] Based on the Revised Bloom’s Taxonomy, this study collected data through experimental methods. Empirical analyses were conducted using two-way analysis of variance, one-way analysis of variance, paired-samples t-test, Kruskal-Wallis test, Spearman correlation analysis, and multiple linear regression. Focusing on user cognition in the process of multi-modal information search under the AIGC context, the study proceeded along two lines of inquiry.First, it examined the effects of perceived task difficulty, task type, task context, and familiarity on cognition. Second, it investigated the influence of cognition on multi-modal search behavior and outcome satisfaction. Drawing on cognitive psychology theories, the study further analyzed the underlying generative mechanisms and constructed a cognitive process model linking multimodal information stimuli to outcome satisfaction. [Results/Conclusion] First, the perceived difficulty of image tasks is lower than that of video and music tasks. In image search tasks, the cognitive dimensions of remembering, applying, and analyzing are significantly higher than those in video and music tasks. Task context has no significant effect on cognition, while familiarity with image and music tasks is positively correlated with users’ cognition. Second, applying and evaluating positively affect the number of searches, while creating is more associated with long-term search behavior. In addition, understanding, applying, and analyzing positively affect outcome satisfaction.

[Purpose/Significance] International Artificial Intelligence (AI) alliances serve as crucial platforms for promoting the integration of artificial intelligence resources and fostering collaborative innovation. Analyzing their systemic structures can provide strategic references for the development of similar alliances in China. [Method/Process] Drawing on the triadic structural model of information ecology theory, this paper constructed an analytical framework from three dimensions: ontology, subject, and environment. By integrating thematic analysis with tools such as ArcGIS and Gephi, it systematically examined the core topics, member composition, and operational foundations of international AI alliances. Furthermore, it extracted the synergistic interaction mechanisms among the elements of the information ecology. [Result/Conclusion] International AI alliances, grounded in technological infrastructure and knowledge resource sharing, have formed six major core themes. The alliance members mainly consist of four types of entities, i.e. enterprises, non-profit organizations, research institutions, and government or public agencies. The institutional environment plays a dual regulatory role in alliance operations, while the social environment may subtly influence their value orientation. The evolution of alliance forms is closely tied to generational leaps in technology and exhibits a “center-periphery” spatial distribution pattern. There are three types of interactions among information ecological elements: horizontal, vertical, and cross-dimensional. Based on these findings, the future development of AI alliances in China should focus on introducing diverse members to enhance subject density, providing targeted policy support to optimize the information environment, and advocating for an open-source culture to facilitate information flow. These efforts will help strengthen China’s competitiveness and discourse power in the global AI arena.

[Purpose/Significance] Obliteration by incorporation (OBI) refers to the phenomenon where methods, theories, and other forms of knowledge cease to be explicitly cited after becoming widely accepted as common knowledge. This process results in an underestimation of the academic influence of such knowledge, posing challenges to the accuracy and validity of academic evaluation. Therefore, in-depth research on this phenomenon and its underlying mechanisms is important to enhance academic assessment frameworks. [Method/Process] Utilizing the arXiv repository as the data source, this study measures the extent of OBI in physics by analyzing the trends of explicit citations, implicit citations, and citation rates. It then calculates semantic evolution indicators of knowledge phrases from multiple dimensions. A multiple regression analysis is applied to explore the associations between semantic evolution and OBI. [Result/Conclusion] The findings reveal a temporal delay between the peaks of explicit and implicit citations. The reduction of semantic variation, coupled with the enrichment of application contexts, facilitates the transition of knowledge phrases into common knowledge, leading to an increased degree of OBI. This study explains the mechanism underlying OBI from the perspective of semantic evolution, thereby contributing to a deeper understanding of OBI and offering valuable insights for refining citation-based academic evaluation systems.

[Purpose/Significance] Against the backdrop of intensifying global technological competition, how to effectively identify and predict underdeveloped technical fields and untapped technical opportunities in technological development remains an urgent problem to be solved. [Method/Process] This paper proposed a national science and technology intelligence analysis framework based on technology control policies texts. By integrating prompt engineering for large language models (LLMs) and generative topographic mapping (GTM), it constructed GTM patent maps and multi-dimensional GTM maps to mine underdeveloped technology nodes and technological gaps. Then, it designed a multi-dimensional evaluation index system and conducted a comprehensive assessment to identify the technological opportunity points with CRITIC method. Ultimately, inverse mapping technology was utilized for technological interpretation. [Result/Conclusion] Taking the lithography technology clauses in the Commercial Control List as an empirical case, the framework identifies 6 underdeveloped key technology nodes and 3 potential technological innovation opportunities in China, and verifies its scientificity and effectiveness. This framework is intended to provide scientific basis for the formulation of science and technology policies, and technology strategy layout of relevant departments.

[Purpose/Significance] In the context of the Fifth Research Paradigm, artificial intelligence(AI) technology is deeply integrated with scientific research scenarios. Focusing on the demand for the intelligent transformation of agricultural research services under the new paradigm, and based on the evolutionary trends of research paradigms and the domain uniqueness of agricultural research, this study aims to construct a theoretical framework and practical pathway for the intelligent development of agricultural research. [Method/Process] First, by combining agricultural research practices with a literature review approach, this paper distinguishes the core differences between traditional agricultural research and AI-driven models, and extracts the characteristic dimensions of the new agricultural research paradigm. Subsequently, it systematically deconstructs an AI-enabled agricultural research service mode from five dimensions: service positioning, target users, service content, operational mechanisms, and supporting elements. Finally, drawing on systems theory, an intelligent service platform with a four-layer architecture(data, model, application, interaction) is proposed, with clearly defined functions and technical logic for each layer. [Result/Conclusion] This study proposes an AI-driven agricultural research service framework with the aim of fostering a corresponding service ecosystem. Furthermore, practical strategies for the service framework are elaborated from the dimensions of institutional support, scenario innovation, model transformation, and talent cultivation, to provide, a theoretical basis and practical guidance for the intelligent upgrading of agricultural research.

[Purpose/Significance] With the frequent occurrence of natural disasters, government Weibo accounts have become an essential way for the public to obtain event-related information and to prevent and respond to risks. This study aims to construct an effective information disclosure quality assessment system, which provides theoretical guidance for the government's evaluation of information disclosure quality and its current status. Based on this, some strategies are proposed to provide references for the government to enhance the quality of information disclosure. [Method/Process] Based on the framing theory, the evaluation framework is constructed from five aspects, including construction subject, textual content, post structure, presentation form, and volume control. Indicators are extracted from related literature. Then, the entropy weight method is used to calculate the weight of indicators, and the TOPSIS and RSR methods are used to evaluate and classify the government Weibo information disclosure quality. Finally, based on the characteristics of information disclosure and the identification of benchmark accounts, a model for high-quality information disclosure of government Weibo is constructed to help the government improve the quality of information disclosure. [Result/Conclusion] The results indicate that the level of quality can be classified into four grades: excellent, good, medium and average. The information disclosure quality of D1 is the highest, while B2 is the lowest, indicating that the quality is not proportional to the level of government. Furthermore, the characteristics of high-quality information disclosure include timely and accurate replies, a high follower count, and appropriate post length.

[Purpose/Significance] Researching a digital twin model for museum exhibition services can provide a theoretical framework for building next-generation digital museum applications that reflect the advantages and characteristics of digital twin technology. It can also offer practical guidance for museums to achieve high-quality digital transformation. [Method/Process] First, this paper analyzes the main features of museum exhibit services based on digital twin technology and their distinctions from traditional systems. Second, it proposes a foundational conceptual model for multi-granularity digital twins, summarizing the structure and characteristics of the digital twin environment from a multi-granularity information perspective. Then, it presents a digital twin model for museum exhibit services, with information recommendation services and functions for organizing display content as its core components. Finally, using the example of touring the Palace Museum, it introduces the implementation mechanism of the digital twin process for related cultural relic exhibits. [Result/Conclusion] From the perspective of multi-granularity information modeling, this paper proposes a digital twin model for museum exhibit services and discusses the design mechanisms of its core functions, providing a reference for the construction of digital museums in the new era.

[Purpose/Significance] Characterized by cloud services, intelligent data integration, and microservice mechanisms, next-generation library services platforms offer a critical solution to the challenges faced by regional university library consortia(RULCs), including resource decentralization and diminishing collaborative efficiency. [Method/Process] This study systematically traces the technological evolution of library services platforms and the development trajectory of RULCs, thereby revealing the core bottlenecks in the transformation toward knowledge services within traditional consortium platforms: closed architectures lead to inefficiencies in digital services; heterogeneous systems cause data format conflicts; loosely coupled cooperation models restrict deep business-process synergy; and closed ecosystems result in data silos. [Result/Conclusion] The study demonstrates that next-generation services platforms can significantly enhance the intelligent service capabilities of RULCs through architectural innovation, resource integration, process re-engineering, and consortium ecological collaboration. Based on this, the research proposes a migration stage model for consortium-level regional university libraries toward next-generation services platforms, promoting the leapfrogging of domestic RULCs from “paper-based resource consortia” to “intelligent service symbiosis.”

[Purpose/Significance] A trusted data space is an emerging infrastructure for the circulation and utilization of data. Exploring the construction path of a trusted data space for the value co-creation of open public data is of great significance for promoting the open development and the circulation and utilization of public data resources. [Method/Process] Based on the logic of the trusted data space empowering the co-creation of open public data value, the social network analysis method was used to evaluate the level of value co-creation and point out the challenges of spatial empowerment. Subsequently, the DART model(Dialogue, Access, Risk, Transparency) for value co-creation was introduced to outline the construction pathways of trusted data spaces, proposing multiple spatial development strategies to enhance co-creation effectiveness. [Result/Conclusion] From the dual logical analysis of theory and practice, space construction can facilitate value co-creation. Due to the low degree of subject aggregation, reliance on individual node driving and other factors, the level of open public data value co-creation needs to be improved. The trusted data space can be constructed from the four aspects of dialogue, access, risk, and transparency by providing a multilateral dialogue platform with centralized resources and diversified access channels with linkage of development paths, accurately enabling the development of open public data value co-creation.

[Purpose/Significance] This study aims to reveal the influencing factors and pathways of algorithmic resistance behavior of mobile short-video users, analyze the dynamic evolution mechanism of cognition and emotion, and provide theoretical basis and countermeasures for the construction of a human-machine collaborative algorithmic governance framework. [Method/Process] The article preliminarily synthesized influencing factors using meta-ethnography and employed the triangular fuzzy number DEMATEL to analyze the causal relationships among these factors. Following this, it utilized the DANP method to quantify the weights of the influencing factors to determine their significance. Subsequently, it employed the AISM method to construct a hierarchical adversarial topological network and ultimately integrated the dual system theory to deconstruct the mechanisms of resistance behavior. [Result/Conclusion] The study identifies 22 influencing factors related to information quality. Based on the dual-system theory, it reveals the impulsive path dominated by System I, the reflective path governed by System II, and the dynamic game mechanism between the two. It elucidates the complex interactions among context, cognition, and emotion in the algorithmic resistance behavior of mobile short-video users. Finally, it proposes management strategies for this behavior from three perspectives: regulatory systems, algorithm design, and literacy cultivation.

[Purpose/Significance] This study explores the pathways, challenges, and development strategies of AI digital humans empowering library service innovation, aiming to provide theoretical support and practical guidance for the intelligent upgrading and smart service system construction of libraries, and to promote the innovative development of smart libraries. [Method/Process] Based on a comprehensive literature review, this study first reviewed the definition, characteristics, and technical architecture of AI digital humans, clarifying their multidimensional value in library services. Second, from the perspectives of library service categories and diverse user needs, it identified the logical path through which AI digital humans are embedded in library services. Finally, by analyzing the challenges libraries faced during the pre-introduction and post-introduction of AI digital humans, the study proposed corresponding development strategies for AI-driven library service innovation. [Result/Conclusion] The findings reveal that the logical path of AI digital human integration into library services comprises four progressive stages: classification of AI-enabled library service categories, configuration of intelligent cores, construction of role types, and adaptation and evolution of service scenarios and capabilities, demonstrating a transition from functional integration to contextual fusion. In terms of service scenarios, AI digital human applications can be categorized into six typical contexts: user consultation and reception, navigation and spatial guidance, academic research and knowledge discovery, teaching and tutoring, companionship and psychological interaction, and brand communication and cultural engagement, each with distinctive service emphases. In terms of development strategies, the study proposes three key approaches, promoting modular development, consolidating data foundations, and advancing pilot implementation, to facilitate the deep integration of AI digital humans into library services.

[Purpose/Significance] This study systematically reviews the current development of government service robots, explores multidimensional factors influencing user preference, and provides theoretical and empirical foundations for establishing user-friendly and efficient government service systems. [Method/Process] Based on the behavioral reasoning theory, the mind perception theory, status quo bias theory, and procedural fairness theory, it constructs a multidimensional and higher-order research model of user preference for government service robots. This model elucidates the formation mechanism of user preference through dual pathways of acceptance and rejection toward government service robots. [Result/Conclusion] The results confirm the significant impact of values on reasons and attitudes, as well as the impact of reasons on attitudes and user preference. Functional values and public values are key in shaping values. Competence is an important factor to form the user acceptance reasons. And perceived risk is the core factor to form the user rejection reasons.

[Purpose/Significance] This study aims to mitigate data conflicts caused by time lags, enhance semantic complementarity and accuracy of multi-source data, and improve the synergistic effect of knowledge fusion. [Method/Process] Focusing on topic time lag, this study incorporates the data source as a covariate in the structural topic model (STM). The direction and duration of the lag are measured by analyzing topic content, evolutionary trends, and similarity. By combining time lag and semantic similarity, the knowledge element semantics are separated into three categories: non-lagging semantics, lagging-similar semantics, and lagging-dissimilar semantics. And semantic separation and fusion strategies are proposed. Furthermore, an implementation framework for a multi-source data integration knowledge fusion model is constructed. This framework is applied and validated using artificial intelligence domain data from NSF and WoS. [Result/Conclusion] The proposed approach of local time lag processing mitigates their negative impact on data fusion while maintaining the characteristics of the data source. It is the basis for designing topic time lag measurement and knowledge element similarity fusion strategies. The calculation reveals that topic time lags of NSF and WoS data manifest as both leads and lags, and exhibit different lag directions and durations on different topics. The fusion of knowledge element semantics significantly reduces semantic redundancy and noise, achieves alignment and deduplication of divergent semantics in multi-source data, while preserving the integrity of semantic information. The proposed workflow of “Time Lag Measurement→Semantic Separation→Semantic Fusion” can effectively support the implementation of multi-source data integration knowledge fusion model.

[Purpose/Significance] In light of the weakness in theoretical research and the lack of replicable operational paradigms for data value realization in this field, this study is to uncover the driving forces and evolutionary paths of data value realization, providing theoretical insight for the efficient utilization and precise empowerment of smart elderly care data. [Method/Process] Adopting a multiple case study approach, the research selects Y Community in Wuhou District, Chengdu, and X Home-Based Elderly Care Service Center in Xinwu District, Wuxi, as case sites. Applying grounded theory methodology, the study conducts three levels of coding on the case materials and constructs a mechanism model of data value release in community-home-based smart elderly care. [Results/Conclusion] The findings reveal that the release of data value in community-home-based smart elderly care is primarily driven by three core forces: stakeholder participation, technological embedding, and institutional guidance. It evolves through a four-stage path—value perception, value activation, value realization, and value emergence—forming a mechanism model of data value release structured around the main logic of “core motivation, path evolution, and continuous release.” This model offers theoretical guidance and methodological reference for fully unlocking the value of elderly care data and optimizing community-home-based smart elderly care services.

[Purpose/Significance] In an open access environment, public digital cultural resources of considerable value or effectiveness constitute a potential economic ‘substitution’ with non-public digital cultural resources. Exploring the factors influencing the substitutability of public digital cultural resources provides a reference path for enhancing the attractiveness and irreplaceability of public digital cultural resources. [Method/Process] The original corpus was obtained through the semi-structured interview method, and the interview texts were analyzed using the rooting theory and the push-pull theory to summarize the factors influencing the substitutability of public digital cultural resources and construct the push-pull model. [Result/Conclusion] In the push-pull model: (1) the internal environment shapes the public/non-public digital cultural resource push-pull; (2) the external environment regulates the public/non-public digital cultural resource push-pull; (3) the public digital cultural resource pull and the non-public digital cultural resource push weaken the public digital cultural resource substitutability; (4) the public digital cultural resource push and the non-public digital cultural resource pull reinforce enhance public digital cultural resources substitutability; (5) individual user factors regulate public digital cultural resources substitutability through users’ willingness to use.

[Purpose/Significance] In the era of big data, scientific and technological texts exhibit characteristics such as multi-source, multi-type, and multi-structured. Designing topic detection methods to accurately organize scientific and technological information from massive texts has become a supportive task for formulating scientific and technological development strategies, optimizing resource allocation, and promoting scientific and technological innovation. [Method/Process] This paper proposes a neural network model, NNMMFF, based on multi-view feature fusion. The model utilizes a self-supervised training framework to extract feature vectors of scientific and technological texts from three perspectives: network, statistical, and contextual. It uses an embedded “multi-view feature fusion module” to integrate the three feature vectors and obtain fused text features. Finally, the effectiveness of this approach is validated through the topic detection task. [Result/Conclusion] Experiments are conducted on a Chinese dataset from the “National Security” domain (2015–2023). The results show that in NNMMFF, a deep feature-level fusion algorithm using gated multimodal units can effectively fuse feature vectors from the network, statistical, and contextual perspectives. In contrast, shallow feature-level fusion algorithms, such as feature concatenation and feature weighting, perform relatively poorly. The findings suggest a clear performance hierarchy for topic detection in multi-source scientific and technological texts: three-view fusion outperforms two-view fusion, which in turn outperforms a single view.

[Purpose/Significance] This study aims to enhance knowledge fusion at the user level, enabling a more systematic and feasible approach to embedding user needs into knowledge services. [Method/Process] Based on the thematic, collaborative, and hybrid relationships between author knowledge elements and semantic knowledge elements, this paper designs four recommendation algorithms within a neural network framework. These algorithms are categorized into three types: content-based, collaborative filtering, and hybrid recommendation. The dataset is divided into training set, test set, and calibration set in a 7:2:1 ratio. An implementation framework for a knowledge fusion model to mine user needs is constructed. The artificial intelligence domain data in WoS database is selected for implementation comparison and result validation. [Result/Conclusion] This paper proposes four knowledge recommendation models: content-based(C), collaborative filtering(CF), content-based collaborative filtering(CCF), and collaborative filtering content-based recommendation(CFC). The outputs from these models can serve as an auxiliary representation for user needs, support the implementation of the knowledge fusion model, and validate the feasibility of the “semantic supplementation-semantic recommendation” path. In addition, the CFC recommendation model achieves significantly higher accuracy than other models in both content evaluation and quantitative evaluation.

[Purpose/Significance] Constructing an industrial technology chain risk measurement method and systematically evaluating the risk evolution trajectory over a long period can provide a methodological reference for breaking through the “bottleneck” dilemma of key core technologies. [Method/Process] Starting from the connotation characteristics of the industrial technology chain and technology gap theory, this study constructed a three-dimensional risk framework of “technology level-technology value-technology complexity”(LVC), and proposed a competitiveness index and risk index measurement method integrating Entropy Weight-TOPSIS based on this. It quantitatively identified the advantages and disadvantages of each link of the technology subjects, and revealed the risk differences and time-series evolution trajectories of different subjects from the perspective of the overall technology chain and technology links. [Result/Conclusion] Taking the “New Energy Vehicle” industry as an example, the empirical results show that the measurement method based on the LVC three-dimensional risk framework can significantly improve the dimensional completeness and decision-making support of the industrial technology chain risk assessment. This study expands and deepens the theory and measurement methods of industrial chain risks, and provides scientific support for the optimization of industrial technology policies, forward-looking technology layouts and technology chain security governance.

[Purpose/Significance] This study integrates indicator features reflecting patent value with semantic features derived from patent texts and leverages the Sparrow Search Algorithm(SSA) to address two major challenges in the identification of potential high-value patents: incomplete feature representation and the complexity of parameter optimization. [Method/Process] It proposes a potential high-value patent identification model, named SSA-BERT-CNN-BiLSTM(SSA-BCB), which incorporates both semantic and indicator features. Semantic features are extracted from patent texts using the BERT model, while indicator features are mined from co-occurrence networks. These heterogeneous features are fused via vector concatenation. The CNN-BiLSTM model parameters are subsequently optimized through SSA to enhance classification performance. [Result/Conclusion] Experimental evaluation demonstrates that the proposed SSA-BCB model achieves a 4.9% improvement in the F1-score over benchmark models, with the area under the ROC curve(AUC) reaching 0.927. Furthermore, the integration of an attention mechanism yields an additional average performance gain of 4.2%. These findings confirm that SSA effectively enhances the identification of potential high-value patents, that the complexity of semantic relationships can substantially influence model performance, and that the combination of optimized parameters with attention mechanisms can further improve identification accuracy.

[Purpose/Significance] From the perspective of technology readiness improvement, this study adopts a three-phase evolution model of application-oriented basic research in chemical industry, dividing the paper citation network chronologically into three phases. By examining the structural characteristics and evolutionary patterns of the network from multiple angles, this article reveals the supporting mechanisms of application-oriented basic research for emerging technology directions, aiming to optimize technological innovation management and promote technological progress. [Method/Process] Taking lithium iron phosphate(LFP) battery technology as an example, this article conducted an empirical analysis through dynamic growth cumulative network topology analysis, community clustering of time-sliced subnets, cross-phase bipartite network knowledge flow analysis, and identification of three-phase paper characteristics in the entire network. [Result/Conclusion] The citation network structure of application-oriented basic research papers exhibits self-similarity, but the undirected self-similar structure is constrained by directed connections, which may impede technological innovation. Papers from the “technology integration and systematization phase” act as bridges connecting the “core technical research phase” and the “technological upgrading phase”. “Knowledge backtracking” helps overcome bottlenecks in technological readiness improvement. Research focuses across the three phases exhibit a “concentration-expansion-concentration” trend, with cross-phase citation patterns breaking the preferential attachment(rich-get-richer) mechanism. This study proposes network intervention strategies to foster technological innovation, emphasizing the importance of monitoring emerging core nodes with growing control and influence to preemptively address potential industrial restructuring risks. The findings provide critical decision-making references for optimizing R&D resource allocation and advancing the readiness of emerging technology directions.

[Purpose/Significance] Investigating the impact mechanism of user concerns on information suppression behavior in online health communities aims to provide a theoretical basis for the “mitigating suppression and increasing the flow” transformation, and to promote the dissemination of health information. [Method/Process] Grounded in the cognitive-affective personality system theory, and combined with protective motivation theory, self-monitoring theory, face theory, and regulatory focus theory, this study formulates hypotheses regarding the paths among latent variables, such as privacy concerns, evaluation apprehension, and prevention focus. A model of the impact mechanism of privacy concerns on information suppression behavior in online health communities is constructed and empirically tested. [Result/Conclusion] In online health communities, privacy concerns, evaluation apprehension, and prevention focus all have a significant positive effect on user information suppression behavior. Evaluation apprehension and prevention focus play a sequential mediating role in the relationship between privacy concerns and information suppression behavior. User privacy concerns have no significant effect on prevention focus. Drawing on the research findings, targeted intervention strategies are proposed to regulate information suppression behavior in online health communities, including strengthening privacy protection settings, improving community evaluation mechanisms, and refining the classification of personality traits.

[Purpose/Significance] Research on the concept of information makes significant contributions to the academic discourse system. From the perspective of information theory, the classical definition is that information is something that can be used to remove uncertainty. Research on the origin and evolution of the definition can reveal the innovative contributions of Chinese scholars to the idea of “uncertainty” in the conceptual history of information, which inspires us to value subjectivity and originality in basic research. [Method/Process] The origin of the concept of information was identified through textual research. Initially, the Chinese literature on the origin of this definition was reviewed. Subsequently, further textual research was conducted on the relevant primary sources in other languages. Ultimately, the true origin of the concept was traced to the early research by Chinese scholars in information theory and information science. [Result/Conclusion] This definition is widely believed to have been proposed by Claude Elwood Shannon(1916-2001), the founder of information theory, in his 1948 work, A Mathematical Theory of Communication, but there is no such statement in his original text. This definition is a creative interpretation of Shannon's information theory by Chinese information scientist Zhong Yixin. It emerged during the period when the “San Lun”(i.e., general systems theory, cybernetics, and information theory) was widely studied in China around the 1980s, and was first published in Zhong Yixin's article The Present and Future of Information Science in 1978. It was widely disseminated due to the significant academic impact of Zhong's book XINXI KEXUE YUANLI(i.e., Principles of Information Science) in 1988, but its content has been misinterpreted. This definition reflects the theoretical connotation and local value of the meta-concept of information science, constituting the ideological core of the academic discourse system of information resource management in China.