Academic Papers Recommendation Based On Deep Learning

Academic papers,as one of the most important academic resources of scientific researchers,play a vital role in the whole scientific research process.With the explosive growth of the academic papers in era of internnet world,most researchers are inevitable to face the increasingly serious problem of "paper information overloading".In response to this problem,some researchers proposed a service called academic papers recommendation..Academic paper recommendation is considered to be an effective way to alleviate the status of papers overwhelming,which can provide a customized papers recommendation service to users.However,traditional recommendation methods have some unpleasant disadvantages at different espects,and cannot effectively generate paper recommendation results.It is urgent to use a technology to enhance the traditional academic paper recommendation technology and improve the recommendation effect and satisfaction.As a branch of machine learning,deep learning technology has made great strides in the fields of natural language processing,image recognition and speech synthesis recently.People have paid much attention to this technology.Thus,if we can apply deep learning technology into the context of academic papers recommendation,combine deep learning `s ability to learn deep and robust features,more effective recommendation results can be generated.This paper focuses on the research of academic papers recommendation,and the main work includes the following aspects:(1)Through the literature,this paper makes a review of the current state-of-arts methodologies.It is found that traditional recommendation methods usually have cold start and data sparse problems,and cannot effectively predict user implicit scores.Thus,the focusing problem of the research is established.Then,a summary of the commonly used recommendation methods,evaluation metrics are presented.(2)Based on the previous research,this paper proposes a novel model based on deep learning techniques called NHPR(Neural Hybrid Paper Recommending).The model fuses the traditional hybrid recommendation with the recurrent neural network in deep learning,which can effectively capture the sequence information in the title and abstract of the paper,generate robust paper project hidden vector,which can improve the prediction accuracy of users implicit scores(3)The model is implemented with PyTorch framework and model evaluation is performed on realworld datasets.Results show that compared with traditional latent factor model MF,the proposed model can overperforms about 26.83% and 46.89% on RMSE repectively.Also,In addition,in order to demonstrate the actual recommendation effect of the model,this paper also made a paper recommendation analysis for random users,and achieved satisfactory results.