The Discovery And Evolution Of Hot Academic Topics

With the wide application of Internet technology in life,the media of academic papers is also evolving.The number of papers that researchers face every day has an explosive growth,which makes researchers unable to study every academic papers.Therefore,how to use topic discovery technology to extract the topic accurately is an urgent problem to be solved.At present,LDA model is commonly used in the field of topic discovery,but LDA model has many problems,such as not using semantic information,unclear meaning of topic representation words and large number of topic extraction.However,other academic topic discovery methods,such as social network method,are complex and time-consuming.Therefore,this paper improves the traditional LDA Algorithm with the characteristics of academic research hotspots,and verify the model effect in the field of spatial econometric model.In this paper,the traditional LDA model is taken as the research object.Firstly,aiming at the problem of no difference in the importance of words in LDA Algorithm,the concept of heat is introduced,and the HLDA algorithm is proposed,which combines the heat of academic papers with TF-IDF to distribute words unevenly,so as to highlight the meaningful vocabulary and change the topic distribution.Secondly,in order to further improve the accuracy of the algorithm and solve the drawback of LDA model without considering semantic information,the K-HLDA algorithm is proposed.Word2 vec is introduced into the K-means clustering process.The probability distribution of topic-word is expressed as word vector of topic,and then K-means clustering is performed to incorporate semantic information and merge redundant topics to improve topic discovery accuracy Degree.Finally,in order to evaluate the applicability of HLDA algorithm and K-HLDA algorithm,empirical analysis is carried out in the related fields of "spatial metrology".Based on the academic topic discovery,the correlation analysis of academic topics in different time segments is carried out,such as topic content evolution analysis,topic intensity evolution analysis and topic life cycle evolution analysis.The conclusions of this paper are as follows: firstly,compared with LDA Algorithm,HLDA algorithm makes the key words more prominent and changes the topic distribution,while K-HLDA algorithm solves the problem that the number of topics in traditional LDA model and HLDA model is too large and has redundant topics,and improves the accuracy of academic topic discovery by introducing semantic information and secondary clustering;Secondly,HLDA algorithm and K-HLDA algorithm are clear and easy to understand in theory.Compared with graph model,topic extraction is relatively simple,and topic heat can be directly obtained by introducing the concept of heat,and there is no need to determine the subject membership of documents;Thirdly,K-HLDA algorithm is applicable in the field of hot academic topic discovery,and it is applied to the topic evolution analysis of the related fields of spatial metrology.It is found that academic topics have low content similarity in the theoretical dimension,and it is difficult to generate evolution relationship.On the contrary,there is a certain evolution relationship in the application dimension.While research methods may not change in different time segments,the research object changes according to the national economic policy.