Topic Recognition Of Policy Texts Based On Dependency Parsing

With the improvement of government digitization and the continuous expansion of the scale of policy texts,the key issue of policy information reuse is how to make full use of the network volume of policy texts to obtain effective information in the current era of big data.At the same time,the application and expansion of literature metrology,deep learning,text mining and other technologies provide a variety of quantitative analysis methods and ideas for large-scale policy text research.Policy text quantification has become the main trend of policy research in recent years.Policy theme recognition is one of the methods of quantitative research on policy texts.Policy themes can reflect the main generalizations of policies in a certain field,so they are widely used in policy analysis.However,the current commonly used topic recognition methods have weak semantic understanding and low comprehensibility.The topic recognition method needs more in-depth semantic information.Dependency parsing,which is characterized by deep semantic information,has attracted much attention in many fields.As a foundation,dependency parsing has been applied by scholars to explore unstructured data such as micro blog,comment data and news data.However,only a few scholars have applied dependency parsing to policy texts,and no scholars have identified the themes of policy texts based on dependency parsing.Therefore,this study focuses on the application of dependency parsing in the subject recognition field of policy texts.This paper constructs a framework of policy text topic recognition based on dependency parsing.Based on the specific phrasing habits of policy text,dependency extraction rules for policy text structure are formulated to achieve fine-grained key phrase structure extraction.The vectorization training of key phrase set was realized by using doc2 vec model.Topic clusters with co-embedding of words,phrases and topics were obtained through top2 vec modeling.Finally,subject words and topic phrases are obtained according to the topic vector.This paper selects China’s policies on science and technology talents from 1978 to2022 as the experimental data set.The subject recognition method proposed in this paper is used for experiments.Combined with subject words and topic phrases,the topic semantic is interpreted.The experimental results are compared with the results without dependency parsing and the results of other topic models.The experimental results show strong intelligibility and semantic interpretation.In addition,This paper identifies the overall and phased theme of China’s science and technology talent policy,and interprets and analyzes the theme.The representative policy theme of each stage is effectively analyzed.Based on this experiment,the effectiveness and feasibility of the proposed method are verified.The policy text topic recognition method proposed in this study realizes the application of dependency parsing in the field of policy text topic recognition.At the same time,the research methods in the field of policy text topic research have been enriched and expanded.It provides a new idea for the analysis of policy text.In addition,there are still shortcomings in this study.This paper summarizes them and puts forward future prospects.