A Scholarly Information Retrieval System Incorporating Recommendation with Semantic Similarity

The scholarly information retrieval systems help us to access a broader range of information. However, the currently available systems do not always ensure data authenticity and real-time. To solve these problems, we propose a Scholar Think Tank System based on an actual scholar database from SCHOLAT to build a high-quality and reliable database. Our system ensures the authenticity and real-time data through data synchronization and manual information collection from scholars not registered to SCHOLAT. In addition, users may be interested in scholars based more on common research content, which reminds us that we need to pay attention to the similarity between scholars. With this inspiration, we have developed a semantic similarity-based scholar recommendation service. We use the pre-training language model SBERT to calculate the similarity between scholars through their profiles and devise an incremental update algorithm to reduce the use of computing resources. Our system has been developed and online.