| Summary: | 碩士 === 國立成功大學 === 工業與資訊管理學系專班 === 101 === With the explosive growth of the emerging popularity of the Internet, it has the trend to publish researching papers on-line. The Institute for Scientific Information (ISI, now Thomas Reuters) has the most widely used online database (Journal Citation Reports, JCR) to provide the most valuable journals. The JCR Science edition contains data over 8,000 journals in science and technology. However, the mass of content available on the Internet arises a problem of information overload and disorientation. Currently, the teachers usually seek the researching papers by keying the keywords from academic search engine. However, the number of selected papers from the searching results is still very large. On the other hand, many teachers get used to using the target-based searching to seek the journal papers on-line. Since they don’t consider the factor of time-varying, it becomes difficult to find the appropriate journal papers to match their current studying.
To overcome above-mentioned problems, we propose a journal recommendation method based on considering time-weighting parameter. Firstly, we utilize the N-gram and term frequency (TF) to classify and categorize words. Secondly, we identify the keywords according to the Java Wikimedia API. Generally, we judge the newer papers to be more important than older papers. Therefore, in order to extract the suitable for researching topics, a time-weighting is set in our method according to time factor of the journal papers. Finally, we make a reference vector (RV) from the set of studying topics of teachers, and utilize the RV to set up the binary vectors of researching topics of teachers and journals. To reduce the complexity of the proposed method, we perform the similarity matching module in binary vector space.
In this thesis, we propose a time-aware journal recommendation method to seek appropriate journals to teachers. Experimental results show that the proposed approach can efficiently improve the accuracy of the recommendation.
|
|---|
