Key Technology Research And Implementation Of Academic Recommendation Based On Community And Reference Network

Along with the rapid development of computer technology, information digitization and network technology, great changes have taken place in academic field. A growing number of researchers find academic paper from network as theoretical support for their research work. However, the academic papers in network have some problems such as content complexity, quality inconformity and so on, which result in that researchers need to spend a lot of time to select and filter papers. This situation directly leads to the waste of research time. In a worse case, due to many papers in the network, researchers spend a lot of time but can’t find suitable science paper. Therefore, according to users’ requirements, how to actively recommend academic papers and authors to users has become a research hotspot.The main work of this paper is to research the key technologies of academic modeling and academic recommendation, in order to recommend academic resource to users effectively. Based on the study about academic modeling methods, community discovery algorithms, academic reference network modeling methods and academic recommendation methods, this paper proposed a novel academic recommendation method based on topic community and two-layer reference network, and implemented the system based on the proposed method to recommend authors and papers to users. Through experiments, this paper determined the related parameters, verify the effect and advantages of the proposed method.The method proposed in this paper includes a new topic model, named as ACTTM (Author Community Topic Time Model), a two-layer reference network which comprising of author and paper layer, user interest model and recommendation list generation method. ACTTM model combines the AT (Author Topic) model and TOT (Topic over Time) model and adds variable representing community. It can model the community information of author, the topic information of community and the time information of topic. The two-layer reference network utilizes ACTTM to add community information to author layer, and designs different attribute values, named as authority value, multiple value and popular value, to describe the property of authors and papers. At the same time, according to users’operation records, the proposed method generates users’interest community list and calculates the attribute value of users. Finally, it recommends authors and theirs academic papers for users based on two-layer reference network and user’s interest model.Based on the personalized academic recommendation system designed and implemented according the above proposed method, we verify that the community discovery based on ACTTM can get better modularity and reflect the dynamic variety of communities with time. The recommend scheme based on the two-layer reference network can effectively reduce the calculation complexity of recommendation algorithm, improve the accuracy of recommendation, and provide more diverse recommendation results.