[Purpose/Significance] Theory is an essential component in the construction and development of the discipline of information science. Organization and Analysis of theories not only help understand the origins and developmental trajectories of the discipline but also predict the development of emerging technologies. Efficient and accurate identification of theoretical entities plays a crucial role in deepening theoretical research. [Method/Process] This paper proposed an information science theory extraction algorithm that collaborates between large and small language models, including modules for enhanced word embedding vectors, sample difficulty assessment, and a theoretical identification model. Initially, the paper used large language models to pre-identify theoretical entities. These pre-identified entities, combined with the original word embeddings, formed the enhanced word embeddings. The training process of domain-specific small models was optimized through these enhanced word embedding vectors. Additionally, the paper used large language models to assess the difficulty of samples and adjusts training strategies accordingly to improve model performance. The proposed algorithm fully integrated the large language models' powerful semantic understanding capabilities and the professionalism of domain-specific small models. [Result/Conclusion] Experiments conducted on a dataset for the extraction of theoretical entities in information science show that the algorithm proposed in this paper effectively improves the performance of theoretical entity extraction, achieving the best results in the metrics of precision, recall, and F1 score.
[Purpose/Significance] Current mainstream recommendation algorithms based on knowledge graph mainly focus on mining and utilizing item-side knowledge, paying less attention to user-side auxiliary information, leading to issues such as sparse user data and insufficient mining depth. [Method/Process] Addressing the structural and feature differences between user-side and item-side auxiliary information, the study proposed a dual-graph attention recommendation model that integrated social relationships and knowledge graphs. Firstly, the user social network graph and item knowledge graph were separately fused with the user-item interaction graph to obtain the user social relationship collaborative graph and item collaborative graph. Secondly, the study used a dual-graph attention network to process these two knowledge graphs separately, extracting different user and item feature vectors. Then, through the attention mechanism, the extracted user and item feature vectors were merged. Finally, the interaction probability of users and items was calculated using inner product operation for recommendation. [Result/Conclusion] The study conductes experiments on the Last-FM and Douban datasets, demonstrating that the model outperforms other baseline models on various datasets.
[Purpose/Significance] In order to effectively solve the dependence and limitation of primitive learners and integrated models on monomorphic specific patterns, this paper expands the classifier into one that can adapt to polymorphic mixed patterns by increasing the observation granularity, with a view to enhancing the model comprehension and classification ability. [Method/Process] In this paper, the study replaced the original features with concept sets, introduced the concepts of mutually exclusive concept sets and orthogonal sample sets, separated, generalized and fused the samples, constructed a concept lattice fusion model and evaluated the model in four aspects: model traits, model capability, model quality and overfitting. [Result/Conclusion] The measurement results with a sample set of 23971 Amazon reviews show that the concept lattice fusion model has a greater improvement in accuracy, stability, and anti-interference, and the model evaluation results indicate that the model has better intrinsic qualities.
[Purpose/Significance] The study aims to test the feasibility and application potential of GPT in the task of identifying the sentence function categories of abstracts, providing a reference for building high-quality structured data based on generative large language models. [Method/Process] The paper used single-round dialogue and zero-shot prompts based on GPT 4.0, Qwen 1.5 and ERNIE 4.0 to perform the category identification task of structure and function. Different test subsets were constructed according to domain, language and time range of the publication. Then, the P, R and F1 values, and accuracy were used as evaluation indicators. And the single-factor analysis of variance was used to measure the different performance between subsets. [Results/Conclusion] The outputs of large language model exceeded the categories restriction in the prompts. However, the high proportion of outputs that meet the prompts shows that using generative models to solve discriminative tasks is basically feasible. Different large language models have different performances. Some indicators of GPT 4.0 and ENRIE 1.5 are significant at the 0.01-level, while others are not. The indicators include: all indicators of the samples in different categories of structural function, the R and accuracy of samples in different fields, and the P and the F1 value of samples in different languages. In the future, when building intelligent intelligence services based on generative large language models, we should focus on the controllability of output results, domain adaptability, etc.
[Purpose/Significance] The study aims to construct a similarity computation model specifically applicable to TCM ancient texts, solve the problems of difficulty in semantic characterization and the high cost of data annotation of BERT on TCM ancient texts. [Method/Process] On the basis of incremental pre-training of multiple models, this study used generative AI to generate all task data, and combined SimCSE method to compare the effects of different training methods, pre-training models, positive and negative sample construction methods, and positive sample mixing strategies. [Result/Conclusion] The research results show that the performance of unsupervised learning models is generally low, and the introduction of positive and negative samples generated by AI significantly improves the performance. On the training set composed of semantically different and low similarity negative samples constructed using AI, and mixed with positive samples constructed using AI assisted synonym replacement method, the TCM Gujiroberta model showed the best performance, reaching 90.9%. In addition, selecting negative samples with low similarity and randomly mixing datasets with different types of positive samples can further improve model performance. This study designs a SimCSE similarity calculation model that integrates knowledge of traditional Chinese medicine ancient books in a zero sample scenario, which can provide support for the research and application of ancient books. In the future, further optimization would be considered in terms of dataset construction strategies.
[Purpose/Significance] This paper focuses on investigating the causes and mechanisms behind users' intention to avoid artificial intelligence-generated content (AIGC)information, aiming to provide theoretical guidance for user information behavior, optimization of AIGC technology, and iteration of AIGC products. [Method/Process] Drawing upon the C-A-C framework, this study constructed a theory model to examine the influence of controllable degree of information, transparency of information, and AI anxiety on information avoidance.The study collected sample data through virtual experimental design, and analyzed data using partial least square structural equation modeling (PLS-SEM). [Result/Conclusion] The findings reveal significant differences in users' avoidance intention towards AIGC information based on its transparency and controllability.Additionally, AI anxiety exhibits a positive relationship with information avoidance behavior.Notably, it is important to acknowledge that AI identity threat positively moderates the relationship between AIGC information uncertainty and AI anxiety.
[Purpose/Significance] The study aims to analyze the complicated relationship among the factors influencing the adoption intention of AIGC users under the scientific research scenario, and analyzes whether there is a single factor that has a necessary effect on the adoption intention of users, it provides theoretical and practical reference for the popularization and application of AIGC in scientific research scenarios. [Method/Process] The necessary condition analysis and fuzzy set qualitative comparative analysis were used to analyze 522 questionnaire data to explore the necessity of single factor and the adequacy of configuration path. [Result/Conclusion] The study finds that there is no single factor that has a necessary effect on the adoption intention of AIGC users, there are 11 high adoption intention paths and 4 low adoption intention paths in the influencing factors configuration paths.Among them, perceived quality, perceived usefulness, perceived ease of use, perceived task technology matching, perceived trust, social influence, and user expectation all play an important role in the configuration path as the core condition or edge condition.
[Purpose/Significance] In the era of AI, artificial intelligence generated content (AIGC)has gradually become a new tool and core content in the field of creation, affecting the creative process and effectiveness of creators.Exploring the impact of AIGC on creators' job crafting could provide specific reference for improving human-AI collaborative creation level and human-AI relationship. [Purpose/Significance] Two experimental studies were conducted on Credom, and the influence mechanism of AIGC-driven job crafting was analyzed by referring to CAT model, and the dimensional framework of dual features of AIGC. [Purpose/Significance] The study found that AIGC has a positive impact on creators' job crafting, in the specific influence mechanism, the perceived entertainment of content has a promoting effect on emotional attachment, and the perceived usefulness of technology has a positive impact on self-efficacy.At the same time, self-efficacy and emotional attachment will further affect job crafting, and cognitive crafting has a significant mediating role in the process of driving task crafting and relationship crafting.This study expands the research content of AIGC in the field of human-AI interaction, and the research conclusions have a marginal contribution to deepening the theoretical relationship between AI and creator work.
[Purpose/Significance] The healthy development of integrated media platforms in the era of big data is an essential support for a strong Internet nation and a crucial component of the digital economy's growth.Research on the information ecosystem within the integrated media environment holds significant value. [Method/Process] Relying on the existing network information ecosystem, this paper adopted the perspective of complex systems to construct an information ecosystem model for integrated media in the era of big data.Through the analysis of internal element correlations, it elucidated the operational mechanism of the model and clarified the evolutionary mechanism under the coordination of external factors.This revealed the rationality of analyzing the development of integrated media platforms using information ecology theory. Simultaneously, it utilized the proposed information ecology theory of integrated media in the era of big data to analyze two well-known integrated media platforms, "Toutiao"and"Qutoutiao".By comparison, it revealed the reasons behind the phenomenal rise and fall of the"Qutoutiao"platform. [Result/Conclusion] The research findings provide a new perspective and innovative ideas to some extent for the construction of the information ecosystem of integrated media in the era of big data.
[Purpose/Significance] The revitalization of Northeast China in the new era depends on the strong support of scientific and technological talents, and the policies for these talents are essential institutional guarantees for promoting scientific and technological innovation and talent development.This study aims to accurately grasp the framework structure and dynamic characteristics of regional sci-tech talent policies through systematic quantitative analysis, providing decision-making references for deepening the supply-side reform of policies. [Method/Process] Taking the sci-tech talent policy texts of Heilongjiang, Jilin, and Liaoning provinces as research objects, this study employed the BERTopic model to conduct topic identification, keyword extraction, similarity calculation, etc.On this basis, it carried out longitudinal and horizontal comparisons of policy themes among the three provinces and conducted a comparative analysis with the policies of developed provinces such as Guangdong, Jiangsu, Zhejiang, and Shandong. [Result/Conclusion] The sci-tech talent policies of the three Northeast provinces have formed a system with talent introduction, cultivation, utilization, evaluation, incentives, and services as the main lines.However, compared with developed provinces, there are still deficiencies in policy supply, demand traction, and targeting.Based on this, the study proposes that the three Northeast provinces should enhance the systematicity, precision, and timeliness of talent policies based on regional realities, focus on key issues such as policy supply, demand traction, and regional characteristics, and continue to make efforts in the integration, optimization, and innovative development of talent policies.The reform of talent policies in the new era should lead and guarantee the comprehensive revitalization and high-quality development of Northeast China.
[Purpose/Significance] Given the critical role of researchers' willingness to share data, the responsible behavior of data-sharing platforms, and the regulatory oversight by government departments in enhancing data-sharing efficiency in China. [Method/Process] The study first analyzed the game-theoretic relationships among key stakeholders in scientific data sharing.An evolutionary game model was then constructed involving researchers, data-sharing platforms, and local government regulatory departments.The model examined the dynamic evolution and equilibrium strategies of stakeholders' decisions under various scenarios.Subsequently, MATLAB simulations were conducted to assess the impact of initial willingness and parameter variations on scientific data-sharing activities. [Result/Conclusion] The findings indicate that the initial strategies and interactions of the three stakeholders have a significant impact on scientific data-sharing activities.The primary factors driving the shift of data-sharing platforms from"non-compliant sharing"to"compliant sharing"include regulatory fines and researchers' compensation capabilities.Factors influencing researchers' strategies include regulatory incentives and the total benefits of data sharing.The key factors affecting regulatory departments' strategies are regulatory costs and accountability from higher-level government authorities.
[Purpose/Significance] The paper explores the influencing factors and mechanism of privacy protection behavior of users in opening government data, so as to guide users to actively participate in privacy protection and improve the efficiency of personal privacy protection. [Method/Process] Based on the MOA theory, the paper constructed a research model of the privacy protection behavior mechanism of users in opening government data, and structural equation model (SEM)and fuzzy set qualitative comparative analysis (fsQCA)were used to analyze investigation data. [Result/Conclusion] The SEM results show that perceived benefits, perceived costs, subjective norms, platform support, privacy literacy and privacy concerns significantly affect the privacy protection behaviors of users in opening government data.The five configuration paths found by fsQCA can be summarized into three types of configurations that trigger users' privacy protection behaviors, namely, platform guiding, privacy literacy driving, and internal and external linkage.Platform support and subjective norms appear as core conditions in the four configuration paths respectively.It shows that the generation of users' privacy protection behavior is the joint result of internal motivation, ability and external opportunity, and the opportunity factor of external environment is the core condition.Based on this, the paper puts forward corresponding suggestions.
[Purpose/Significance] User health empowerment in online health knowledge services is scene-dependent and adaptable.However, existing studies on user health empowerment have limited applicability in this context due to significant differences in interaction mode and autonomy compared to traditional medical scenarios.Therefore, it is crucial to consider these differences when interpreting the connotations and dimensions of user health empowerment. [Method/Process] This study confidently employed a mixed-methods approach to expertly explore the structural dimensions and measurement of user health empowerment in online health knowledge services.Firstly, the study extracted the structural dimensions of user health empowerment from user interview data.Then, a standardized process was assertively followed to develop a measurement scale. [Result/Conclusion] The results clearly demonstrate that online health knowledge services significantly contribute to user health empowerment.This construct is comprised of three secondary constructs(i.e., psychological empowerment, structural empowerment, and resource empowerment)and eight tertiary constructs.Psychological empowerment aligns with the measurement dimensions of user health empowerment in traditional healthcare settings, but with richer connotations.The structural and resource empowerment dimensions reflect the new phenomenon of doctor-patient interactions in the context of online health knowledge services.The scale consists of 27 items and has shown excellent reliability and external validity through exploratory factor analysis, validation factor analysis, and external validity tests.
[Purpose/Significance] The online medical community has a large number of doctors and a limited amount of information available to patients.As a visual symbol, the doctor's profile photo is the first-hand information that patients are exposed to, influencing their decision-making behavior.By analyzing the influence of doctors' profile photo information on patients' decision-making behavior, the study aims to support doctors' self-presentation, patients' decision-making, and platforms' operation optimization in online medical communities. [Method/Process] Taking HaoDf as the research context, the study captured the data of 47704 doctors in representative departments of HaoDf.Based on the self-presentation theory, this study employed regression analysis to analyze the influence of the doctors' profile photo content and equality on the patients' decision-making behavior in the online medical community. [Result/Conclusion] The study finds that besides the richness of doctors' profile photos, the authenticity, professionalism, affinity, focus, and clarity of doctors' profile photos significantly affect patients' decision-making behavior.
