Increasing the Authoritativeness of Literature Recommendation

• Main Authors: Yi-Rong Lin, 林依蓉
• Other Authors: Wan-Shiou Yang
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/03652163140776942391

碩士 === 國立彰化師範大學 === 資訊管理學系所 === 96 === Literature recommender systems have been shown to greatly help users in navigating scientific documents and locating useful information from literature database that match their interests. Content-based approach is one of the main approaches used to build literature recommender systems. Content-based approach typically involves the applications of information retrieval techniques to build user profiles through the content analysis of scientific documents that user currently or historically navigated. Scientific documents that have a high degree of similarity with user’s interest profile are then recommended. Though this approach has been shown certain extent of success, however, it has an intrinsic problem: it looks at a scientific document to evaluate, almost as a piece of text. Because of this intrinsic problem, content-based literature recommender systems may recommend scientific documents that have a high degree of content similarity but may not by themselves be indicative they are authoritative. In this research, in order to the above problem of the content-based approach, we therefore propose two approaches that exploit the citation network of scientific documents to facilitate the automatic construction of a literature recommender system. The performance of the proposed approaches has been evaluated by using empirical data. The experimental results demonstrate that the proposed approaches outperform the traditional content-only approach.