Design And Implementation Of University Professor Recommendation System Based On Text Handling And Cooperation Relationship

With the advent of the Internet era,how to retrieve information efficiently has become a problem.People often face huge amounts of data,but it is difficult to obtain the data they want.Therefore,recommendation systems are now attracting much attention.Different from the product recommendation and social recommendation in commercial companies,the tra-ditional recommendation algorithm is not suitable for this field and it is difficult to achieve good performance.This makes it very inconvenient for those who want to seek professor cooperation in colleges.In this context,this paper improves the traditional recommendation algorithm by studying personalized recommendation algorithms.Based on the particularity of the problem scenario and the data,based on text processing and network mining,combining multiple features to model the data entity,a feature learning algorithm FLTR4 Rec for professor recommendation is proposed to achieve accurate recommendation for college professors.This paper veri-fies the effectiveness of the algorithm on a real dataset,and finally implements a professor recommendation system based on the algorithm.The main content of the paper is as follows:(1)In the processing of text information,given the nature of the data containing a large number of proper nouns in this scene,this article has studied the extraction of Chinese text features.Considering the difficulty of word segmentation,this article gives up the traditional word segmentation followed word vector strategy,and instead directly learns the text vector of character granularity.By introducing the external corpus of the Baidu Encyclopedia,the fine-grained character vector are pre-trained.The visual method checks the quality of the character vectors.(2)The data in this project not only contains text,but also the cooperation relationship be-tween entities.This paper introduces heterogeneous networks to model the data and uses hyper-edge sampling and auto-encoders to process and extract the relationships in the data.By using the relevant technology of network mining,network features in the data can be properly extracted.(3)Based on text features and network features,a multi-feature learning algorithm FLTR4 Rec that can be used in recommendation scenarios is proposed.Using the powerful learning ability of neural networks,the algorithm uses an inter-entity relationship prediction task to learn the text features and network features of each object in the data,and can automatically fuse the two features.Experiments on real data show that FLTR4 Rec performs better than traditional recommendation algorithms and single feature learning algorithms in professor recommendation scenarios.(4)Based on the FLTR4 Rec algorithm,a university professor recommendation system is implemented.This system is mainly implemented using Django.It has the functions of data display,professor recommendation,and data operation.