Research And Implementation Of Recommendation Algorithm In Open Access System

As the developing of open access in scientific and research area,researchers are getting more used to retrieving literature and communicating online.With explosive growth in the number of papers,how to provide researchers with convenient access to papers has become a great challenge for open access systems.The introduction of paper recommendation and expert recommendation provides new approach for researchers to access information.However,traditional recommendation methods mostly only concentrate on text content.In nowadays open access systems,as users'behavioral data are also abundant,taking this into consideration can surely improve recommendation system's performance.In this paper,based on sufficient research of traditional recommendation algorithms,we propose a new hybrid method which combines collaborative filtering algorithm and content-based algorithm.We first use word embedding to compare similarity between papers' content,then use the nearest neighbor model to compare users' behavioral data,and finally recommend concerning papers based on overall consideration of paper content and users' behavior.Based on the results of previous step,we further study and implement an expert recommendation algorithm by comparing the similarity between experts' paper collection.To verify the effectiveness of our algorithms,we compare our solution with many traditional recommendation algorithms on a public available dataset,and the result shows that our hybrid recommendation algorithm performs better than counterparts.We also implemented the function modules in our open access system using the content based part algorithm as users' behavior data not available yet.Finally,we add function modules in the system to provide convenience for researchers to record their papers into our system by either online or offline method.