Research And Implementation Of Research Collaborator Recommendation Based On ResearchGate

With the rapid development of Internet and the wide application of information technology,social network has become an important part in our learning and life.Through social network platform,users can browse information they are interested in and publish their own edited information.Similar to the traditional social network platform Twitter,ResearchGate is a social network website that provides scientific researchers around the world with a platform for research achievement sharing and academic exchanging.Users can follow other users whom they are interested in,and conduct academic exchanges with others on ResearchGate.Therefore,it would be very meaningful if we can help students find mentors in related fields or help other researchers find peers with the same research interests in such an academic environment.Based on the existing researches,this thesis focuses on using researchers' text data and other information published by researchers to build a user interest model and finally a collaborator recommendation system;(1)In the process of processing the text data of the publications,there is a problem that the thematic representation of topic feature vector is not good enough when directly applying the LDA(Latent Dirichlet Allocation)topic model on the text data.This thesis conducts in-depth research on text feature filtering and topic feature extension.Through the introduction of an external knowledge database Wikipedia,we construct lexical feature lists and LDA topic models based on the Wikipedia categories to filter text features and extend topic feature respectively.A topic feature extension model based on Wikipedia is constructed to improve the performance of topic feature vectors.In addition,some validation experiments are carried out on the 20newsgroup and NSF Research Awards Abstracts 1990-2003 datasets.Through comparison,the validity of the topic feature expansion model constructed in this thesis is demonstrated.(2)Build a usable scientific collaborator recommendation system.This thesis crawls the data on ResearchGate by writing a web crawler program to obtain user information,and discusses related technical problems and solutions for them.Then by using the information published by the researchers such as the dissertation text and the following relationship,a user interest model is constructed based on the topic feature extension model.Finally,a collaborator recommendation system is built using the already built user interest model:for students,the collaborator recommendation system can be used to recommend mentors,and for other scientific researchers,the recommendation system can be used to recommend relevant peers.