Research On Paper Recommendation System Based On Deep Neural Network

With the continuous expansion of Internet information,in modern society,most of the scientific research results on the Internet exist in digital form.With the explosive growth of digital information,it will lead to an urgent problem to be solved-information overload.In scientific research fields such as IEEE Xplore and ACM,information overload is an inevitable and urgent problem to be solved.At the same time,the researcher will choose the right publication in the process of publishing the paper,which will be the first key problem to be solved,but the researchers will find that many corresponding journals or conferences are retrieved through the information retrieval system.It is difficult to decide in this situation.The practicality of the recommendation system for solving information overload problems should not be underestimated.Given its widespread adoption in web applications,it has the potential to address the problem of users having too many choices to choose better information.Researchers in computer science or related field are no exception and finding information that truly fits the field of their research in such a wide range of information is not easy.The publication recommendation system can assist researchers in selecting appropriate conferences and journals.Recently,deep learning has attracted considerable interest in computer vision,machine games,bioinformatics,speech recognition,natural language processing,online advertising and other research fields.It is not only because its outstanding performance,but also the unique characteristics of the feature representation of deep learning automatic learning input data.The impact of deep learning on various fields is obvious,and we propose a recommendation algorithm to show that deep learning related techniques are equally effective in recommendation systems.In order to assist the scientific researchers in the computer field to select the appropriate publications in the submission process,we firstly designs a c “A-class” conference,journal abstracts online crawling system to achieve the abstract of scientific papers recommended by the China Computer Federation.Then,we proposed Mask-Deep-LSTM-CNN(Mask-DLC)recommendation algorithm to recommend appropriate journals and conferences.In this thesis,we give the details of how the deep learning algorithm is introduced into the recommendation system.The two most popular algorithms in deep learning are convolutional neural networks and long short-term memory networks.Our Mask-DLC recommendation model based on these two networks.In order to verify the validity of the recommendation model.We compare it with other recommended algorithms based on these two networks and traditional methods.The final experimental results indicate that our proposed Mask-DLC recommendation model is superior to other models.Finally,in order to demonstrate the accuracy of the model and the efficiency of the system,we designed the Mask-DLC paper submission recommendation system based on the proposed algorithm to assist computer researchers to speed up the knowledge broadcasting.