Constructing Precise Profile Of Scholars Automatically

With the rapid development of science and technology,the number of scientific researchers and academic achievements are increasing and the academic data such as papers and journals are also showing a high-speed growth trend,which marks the arrival of the big data era in the academic field.However,the overload of information makes it difficult to find scholars,so it is necessary to mine a structured and precise profiles of scholars.The main purpose of this paper is to construct a precise scholar profile based on academic data using mining technology.The profile includes three aspects which are scholar's basic attributes,scholar's research interest and scholar's academic influence.Based on the scholar profile,this paper will go a further study of scholar recommendation system.The first dimension of precise profile is scholar's basic attributes.This paper identifies scholars' home pages from multi-sourced heterogeneous web pages and extracts scholar's basic descriptions by using information extraction technology.Firstly,we get the first returned page of Google search by using the names of scholars and the information of the institutions that scholars belong to.Secondly,we filtrate home pages of scholars by rule-based method.Thirdly,trigger words and dictionaries are formulated to different rules for extracting scholars' basic attributes.The extracted scholar's basic attributes will include gender,personal photos,email address,position,and nationality.Scholar's research interest is the second dimension of the scholar profile.The paper discovers tags of scholar's research interest based on academic papers.Firstly,this paper adopts different text representing methods,LDA and Doc2 Vec,to show scholars and interest tags respectively.And then,according the cosine similarity between scholars and interest tags,the five tags with the highest similarity are serve as scholars' interest tags.Finally,the result through weighted voting combining the above two methods will be the ultimate.The third dimension of the precise profile is academic influence of scholar.This paper predicts the number of citations of scholars by machine learning method.Firstly,Feature engineering is constructed from academic papers,including statistical features,text content features and network features.And then,whether the total citation is zero or not will be judged by automatic classification method.Furthermore,this paper predicts non-zero citation through regression technology.The experimental result in this paper shows that better results can be achieved by machine learning method to predict scholar's total citation.At last,this paper obtains the precise profile of scholars by merging three above dimensions.Based on this scholar profile,we develop a scholar recommendation system to provide a visualized precise profile except scholar recommendation functions,and also provide services such as scholar database query and paper query etc.In conclusion,this paper studies the scholars precise profiles including basic attributes,research interest and academic influence,which benefits academic field to follow scientific research updates and scientific personnel developments and also helps the arrangement of scientific talents.