Topic Identification, Evolution And Time Lag Analysis Of Multi-source Scientific Literatur

Science and technology(S&T)literature resources are crucial for the dissemination of scientific knowledge.Through a comprehensive analysis of multiple S&T literature sources,we can explore the topics within the field,establish the developmental rules and characteristics of scientific knowledge,and provide ideas for developing S&T policies and planning strategic layout.To investigate the domain of privacy,we gathered data from various sources of S&T literature,such as papers,patents,and fund projects.We utilized the Latent Dirichlet Allocation(LDA)model to identify the research topics of different S&T literature under ten different time windows spanning from 2013 to 2022.Based on the topic identification results,we categorized the privacy topics and gave them appropriate labels.Moreover,we analyzed the evolution of the identified topics,including temporal association evolution and structural characterization analysis.We used the cosine similarity index to construct a similarity matrix of topics to map the topic evolution paths and explore the similarities and differences in the evolution trends of privacy topics.In terms of structural characterization,we constructed multi-dimensional indicators of topic popularity,topic novelty,and topic influence to explore the similarities and differences of cutting-edge emerging topics and hot topics in different S&T literature.Last but not least,we calculated time lags for multi-source S&T literature.We used the Hungarian matching algorithm and Auto-Regressive Distributed Lag Model(ARDL)to calculate the time lag direction and time lag period of patent-paper and fund project-paper from both internal subject and external quantity perspectives.After analyzing the results obtained from the above-mentioned method,six key research conclusions have been identified.Regarding topic identification research,it has been found that:(1)The research topics of S&T literature in different time windows exhibit both uniformity and variability.(2)There are differences in the topical focus of different S&T literature within the same time window.(3)The privacy topics of multi-source S&T literature can be classified into four categories: privacy protection technologies,privacy in application areas,privacy ethics and laws,and privacy behaviors.These four categories can serve as a basis for analyzing the topic evolution and time lag computation in future research.(4)In studies that focus on the temporal associations of privacy protection technologies,it has been found that the topics within this category not only show a strong internal correlation but also have a significant external diffusion and migration.When examining privacy topics within application areas,the topical evolution paths derived from different S&T literature are both similar and distinct.Additionally,the topics within the categories of privacy ethics and laws,as well as privacy behaviors,show a tendency to cross,merge and split with one another in the evolution process.As a result,they absorb other topical content and create multiple evolutionary paths.(5)In terms of analyzing the structural characterization of S&T literature,there are overlaps and differences between hot topics,emerging topics,and important topics.Identifying emerging hot topics can improve the utilization of resources across multiple sources of literature and provide decision support for researchers in related fields.(6)From the perspective of internal subject,knowledge transfer from patents lags behind that of papers by one year,and knowledge transfer from papers lags behind that of fund projects by two years.From the perspective of external quantity,the significant impact of papers on patents takes five years,while the significant impact of fund projects on papers takes two years.These findings are consistent with time lags observed in other fields.This study provides a new idea for the development of more standardized strategies for fusing S&T literature in the future.