[Purpose/Significance] Metaphor understanding depends on specific contexts.However, machines are unable to make inferences based on context when understanding images, making it difficult to grasp the metaphorical meaning behind the images.Constructing a knowledge description framework of image metaphor and proposing strategies for understanding image metaphors will promote understanding of image metaphors. [Method/Process] By reviewing the research of image metaphor understanding, taking complex and abstract psychological health images as an example, the study constructed a knowledge description framework of image metaphor.Based on the framework, the study annotated 351 psychological health images, and proposed four image metaphor understanding strategies based on the contextual relevance and abstraction level. [Result/Conclusion] The results shows that the knowledge description framework of image metaphor consists of five parts: image semantics, image context, metaphorical mapping relationship, metaphor type, and metaphorical meaning.The strategy of understanding metaphors in"direct-concrete"images is association understanding based on image-text matching, the strategy in"direct-abstract"images is direct parsing based on keywords, the strategy in"non direct-concrete"images is indirectly inferring based on semantic association, and the strategy in"non direct-abstract"image is comprehensively understanding based on perceptual similarity.This work provides references for research on machine understanding images.

[Purpose/Significance] The paper aims to generate high-quality synonymous sentences for academic texts utilizing large language models and enhance the performance of natural language inference model through the implementation of semantic enhancement strategies. [Method/Process] Based on the utilization of large language model to generate synonymous sentences for academic texts, the paper proposed a semantic-enhanced natural language inference model, SENLI.The model consisted of a representation module, a semantic enhancement module, and an inference module.Specifically, the representation module was responsible for capturing the semantic features of academic texts and their corresponding synonymous sentences.The semantic enhancement module integrated the semantic features of the synonymous sentences into the original semantic features of the academic texts, thereby obtaining semantic-enhanced features.Finally, the inference module predicted the semantic relationship between pairs of academic texts based on both the original semantic features and the semantic-enhanced features.The study conducted an empirical study by applying the proposed model to the SciTail, SciNLI, and ZwNLI datasets. [Result/Conclusion] The experimental results show that the SENLI model achieves accuracy rates of 95.11%, 79.20%, and 98.43% on the SciTail, SciNLI, and ZwNLI datasets, respectively.Compared to the baseline models, the improvements are at least 1.27%, 1.08%, and 0.92%, demonstrating the effectiveness of the proposed model.The utilization of synonymous sentences generated by large language models for semantic enhancement can enhance the performance of natural language inference model.The research contributes to advancing the field of natural language inference and provides potential technical support for applications such as information retrieval and academic literature mining.

[Purpose/Significance] This study aims to explore the main genres and interaction modes in the practice of online health science communication for older adults and analyze the emotional tendencies of the target audience to optimize the effectiveness of health science communication for this demographic. [Method/Process] Adopting a netnographic approach, this study employed a genre analysis framework to identify the genre types of online health science information for older adults.It also utilized participant observation to examine interaction modes and applied sentiment analysis techniques to assess the emotional responses elicited by different genre types. [Result/Conclusion] The findings indicate that different communication genres exhibit distinct content characteristics and evoke varying emotional tendencies.Diverse interaction modes play a crucial role in enhancing communication effectiveness.The interplay between communication genres and interaction modes provides practical guidance for optimizing genre design and interaction strategies in online health science communication, facilitating broader participation of older adults.

[Purpose/Significance] This study investigates the factors and internal mechanisms influencing users' breakthrough intention regarding information cocoons under intelligent algorithm recommendations in short-video platforms, aiming to provide strategic suggestions for platforms to help users transcend information cocoons. [Method/Process] Drawing upon the Cognition-Affect-Conation (CAC)theoretical framework and the theory of ambivalent attitudes, the study constructed a model examining factors affecting users' breakthrough intention from information cocoons.Empirical analysis was conducted through questionnaire surveys and Partial Least Squares Structural Equation Modeling (PLS-SEM), while multi-group analysis was employed to explore the differential effects across short-video content types. [Result/Conclusion] Perceived control reduces users' breakthrough intention by enhancing flow experience and decreasing psychological reactance.Perceived information homogeneity generates ambivalent attitudes, simultaneously decreasing and increasing breakthrough intention.Perceived information overload intensifies psychological reactance, thereby promoting breakthrough intention.Significant differences exist in users' emotional experiences and breakthrough intention between hedonic and utilitarian video content.

[Purpose/Significance] The layout of strategic emerging industries requires the coordination between basic scientific research and technological innovation activities, and the exploration of the science-technology (S&T)linkage is a crucial pathway for uncovering S&T synergies. [Method/Process] In response to the multi-scale nonlinear associative characteristics of S&T systems, this study constructed an S&T knowledge network in the field of artificial intelligence from the perspective of complex networks and time series mutual representation.The study designed a"network-time series"equivalence conversion method, representing the S&T knowledge network as a nonlinear time series.The paper measured the evolution of S&T synergies using knowledge network distance and sequence synchrony indicators. [Result/Conclusion] The proposed method uncovered the nonlinear dynamic associations within S&T systems.The analysis revealed that the S&T synergy in the field of artificial intelligence became increasingly closer, while due to the effects of scientific research specialization and technological differentiation, the disparity between scientific and technological knowledge structures gradually intensified.

[Purpose/Significance] Disruptive technologies follow unique developmental trajectories, and identifying their lifecycle stages is of significant for technology investment, strategic planning, and policy formulation. To more accurately characterize the lifecycle of disruptive technologies, this study proposes a machine learning-based lifecycle identification method based on the Gartner Hype Cycle. [Method/Process] Specifically, the method constructed a comprehensive measurement system based on three dimensions: academic papers, patents, and user expectations, encompassing a total of 14 key indicators. These indicators analyzed and trained the model, aiming to capture the characteristics of disruptive technologies across various lifecycle stages. The study selected seven representative disruptive technologies, including biochips and cloud computing, for model training, and employing the XGBoost classification model to learn and identify the lifecycle stages as marked by the Gartner Hype Cycle.Then the study selected next-generation artificial intelligence technology for empirical testing to verify the model's applicability and accuracy. [Result/Conclusion] The results indicate that the lifecycle identification method based on the Gartner Hype Cycle effectively identifies the lifecycle stages of disruptive technologies and identify the turning points in the development of next-generation artificial intelligence technology as occurring in 2014, 2017, and 2022. Compared to the traditional S-curve model and inverse S-curve model, the method shows higher sensitivity and accuracy in detecting changes and trends in the early development stages of technology, which aids in better understanding the pulse of technological innovation.

[Purpose/Significance] Is there any difference in title terms between disruptive and continuing innovation patents? What words do they prefer? What are the differences about characteristics between these words? [Method/Process] To solve these problems, this paper selected patents in the field of molecular biology and microbiology. First, the study divided the patents into continuous and disruptive types by using D index, and then displayed two patents of the difference of high frequency words in the title by comparing word frequency, the co-occurrence graph, at last, put forward the innovation type measurement to measure each of the headings of innovative and selection process to distinguish the larger research subject and writing style. [Result/Conclusion] The research shows that there are significant differences between disruptive and continuous innovation patents in terms of title words. Continuous innovation patents often use words with the tendency of gradual improvement and further discovery, such as "Detection" and "with", and often use "Methods" and "motivation" together. The title of disruptive innovation patent is commonly used with some emerging and novel words such as "Human", and "Methods" and "Producing" are often used together. In addition, the innovation type metric can well select the research subject words and writing style words with great innovation difference between the two types of patent titles.

[Purpose/Significance] Eliminating the adverse effects of data element policy uncertainty, promoting the compliant trading and safe circulation of data elements, which hleps to give full play to the multiplier effect of data elements, and accelerate the market-oriented allocation of data elements. [Method/Process] A data element policy uncertainty index was constructed based on data element policy uncertainty-related reports from seven news media in China from 2016 to 2022. The mechanism of policy uncertainty on the development of the city data element market was empirically tested using panel data from 2016 to 2022 for 296 cities in China. Additionally, the analysis was expanded to examine its intrinsic influence mechanism and urban heterogeneity. [Result/Conclusion] The study finds that data element policy uncertainty can significantly promote the development of city data element market and passes the robustness test; the results of the impact mechanism test show that exploring the equilibrium point of financial supply of each production factor, increasing the attraction of talents and promoting the utilization of information resources are the key ways to promote the development of city data element market under uncertain environment. In addition, there are significant differences in the development of the data element market depending on the geographical location, administrative level and digitization level of the city: the impact of data element policy uncertainty is more significant in cities in the east and west, priority cities and cities with high and medium digitization level.

[Purpose/Significance] The cross-border data flow governance policy is an important basis for the governance of a country's cross-border data flow in the context of global digital transformation, and plays an important guiding role in improving the data governance system and safeguarding the security of national data resources. [Method/Process] To analyze Chinese current major cross-border data flow governance policies, this study constructed a framework containing executing subjects-policy tools-policy objectives, revealing the main content, basic logic, and implementation path of Chinese governmental cross-border data governance system. [Result/Conclusion] According to the findings, there is a relatively single main body for the implementation of cross-border data governance policies in China, with less involvement of third-party organizations. Chinese cross-border data governance policies are in an imbalanced state of supply-heavy and demand-light in the policy instrument dimension, with polarized content structure of environmental instruments. Insufficient coordination in the goal dimension, focusing primarily on data security, flow efficiency, and standardized management, while neglecting the goal of personal information rights and interests. It is believed that the policy system should be improved from the standpoints of balanced policy instrument configuration, synergy of multiple subjects, and optimized combination of objectives, in order to promote the safe and orderly flow of data across borders and help realize the value of Chinese national data resources.

[Purpose/Significance] By revealing the dynamic process of medical data sharing system among medical institutions and the dynamic evolution trend of influencing factors, the paper aims to improve the research depth of medical data sharing and provide important basis and suggestions for medical data sharing. [Method/Process] The paper explored into the key factors affecting medical data sharing based on the theory of planned behavior and social capital.A system dynamics model of medical data sharing among medical institutions was developed, followed by simulations and predictions of the data sharing process. [Result/Conclusion] During the first 14 months of the simulation period, the influence of various factors on medical data sharing is ranked as follows: organizational sharing environment > sharing platform=data resources > institutional protection > individual willingness to share, and individual willingness to share grows fastest throughout the simulation cycle.In the first 18 months of the simulation period, the impact of expected benefits on medical data sharing is greater than that of sharing costs; however, after 18.5 months, the situation reverses.After 27.5 months of the simulation period, the insufficiency of data resources and the sharing platform significantly lowers the level of medical data sharing.The influence of law on medical data sharing is consistently higher than that of policy and funding.

[Objective/Significance] Online health communities provide online health services for users, and analyzing the potential information of user comments is of great significance for improving the quality of medical services and optimizing the information construction of healthy communities. [Method/Process] The study proposed an online health community user comment analysis model.Firstly, the LDA topic model was used to dig the topic of patient comments.Next, the classification models were used to classify the patient comments into different topics.Finally, word frequency screening, TF-IDF and SO-PMI method were used to construct a domain sentiment dictionary which was employed to calculate the sentiment score of patient comments in each topics.And comment information of different sentiments were further analyzed. [Result/Conclusion] Empirical research is conducted by analyzing the user comment data of General Grade Three hospital from"Good Doctor Online".According to the information content and rules of the experimental results, relevant reference suggestions for improving medical service and information construction are put forward.

[Purpose/Significance] The paper systematically reviews and analyzes the advancements in conversational systems within the domain of age-friendly design, with the objective of enhancing the theoretical discourse and fostering practical applications.It offers theoretical foundations for the design of conversational systems that address the genuine requirements of older adults, facilitating their integration into the digital society. [Method/Process] The paper employed a systematic literature review method, analyzing data from 96 articles in Chinese, Japanese, and English.It performed a comprehensive analysis of the design motivations, design paradigms, design elements, and evaluation systems of age-friendly conversational systems across four dimensions, established a holistic research framework for the age-friendly design of conversational systems, and proposed four research directions for future study. [Result/Conclusion] The paper, drawing from the interdisciplinary perspectives of Information Science, Information Systems, and Design Science, meticulously examines the principal research accomplishments in the global arena of age-friendly design for conversational systems, thoroughly analyzes the current state, the practical progress and intrinsic logical correlation of the topic and proposes future research and practice directions.

[Purpose/Significance] By systematically reviewing the research literature in the field of intelligent agent role modeling, this paper deconstructs the mechanisms of human-machine collaboration in complex application scenarios, providing theoretical support for the iterative upgrade of intelligent agent design paradigms. [Method/Process] The study employed a meta-synthesis approach to conduct coding analysis on 89 carefully selected Chinese and English literatures, deconstructed the research context of intelligent agents' situational adaptability from three dimensions: role requirements, role creation, and role interaction. [Result/Conclusion] From an integrated perspective of human-ai interaction and situational adaptability design, this paper proposes a three-layer framework: "requirement-defined boundaries(What)→technical implementation concepts(How)→mechanism validation value(Why)"."Requirement-defined boundaries"analyzes role requirements and drives technology selection; "Technical implementation concepts"focus on role creation and personalized design, transforming needs into executable models; "Mechanism validation value"explores role interaction mechanisms, generating feedback data that flows back to the requirement layer.Despite progress in emotional understanding, contextual analysis, and multimodal interaction, current intelligent agents suffer from limitations such as memory constraints, role hallucinations, and role reversals.Future research should center on the deep integration of these three layers of mechanisms, promoting innovations in memory expansion, role alignment, privacy protection, and further advancing the widespread application of intelligent agents.