Research On Review Expert Recommendation Method Based On Text Mining

The emergence of the Internet and the advancement of science and technology have led to a significant increase in the number of applications for scientific and technological projects.The evaluation of the project is very important.Project evaluation often needs to select authoritative experts in related fields from a large number of experts to evaluate and review the project.The traditional method is recommended by the managers from the expert database,which has the problems of time-consuming,wastes manpower and even unfairly distributed due to unfamiliar division of the research fields.Therefore,it is very necessary to study the methods of project review expert recommendation.The core idea of review expert recommendation is to match the project and the expert by using project description documents and expert information description documents,as well as expert historical review and other information.The combination of text mining and recommendation algorithms is used to effectively recommend review experts.This thesis focuses on the key issues such as vector representation models,similarity comparison methods,and recommendation algorithms for projects and experts,and proposes a method for project review expert recommendation based on text mining.This method mainly classifies projects and experts according to their data characteristics and importance,and uses word2 vec and LDA text mining algorithms to build models of projects and experts.The best parameters for model training are confirmed through experiments.The similarity between experts and projects using the similarity measurement method based on weight distribution proposed in this thesis,according to the different contributions of the two types of information to the overall information,the similarity calculation results of word2 vec and LDA are given different weights.The recommendation algorithm proposed in this thesis first uses the restrictive conditions to screen experts,narrow the scope of the experts to be selected,and speed up the calculation efficiency of the algorithm;then uses the weighted similarity comparison method to adopt the content-based recommendation method for recommendation,using the expert model and project model similarity comparison,in which the expert model includes historical review items and expert information,the initial expert recommendation results are obtained;finally,an expert scoring model is established to adjust the recommendation list,and the final recommendation results are obtained.Based on the text mining technology,this thesis proposes an expert review recommendation method.In order to verify that the method is reasonable and effective,a set of comparison experiments is designed to verify.The two sets of comparison methods and the method in this thesis make expert recommendations on the same data set.The experimental results show that under the same experimental conditions,the method in this thesis is significantly higher than the other recommended methods in the F1 value index.This shows that the expert review recommendation method proposed in this thesis can effectively solve the problem of expert recommendation during the project review process and improve the accuracy of the recommendation results.