Topic Analysis of Document Streams Based on Document Arrival Rate and Document Relevance
Chunhua Cui, Hiroyuki Kitagawa · 2004
Topic Analysis of Document Streams Based on Document Arrival Rate and Document Relevance Chunhua CUI Hiroyuki KITAGAWA †Master’s Program in Science and Engineering, University of Tsukuba, Tsukuba, Ibaraki 305-8573, Japan ‡Institute of Information Sciences and Electronics, University of Tsukuba, Tsukuba, Ibaraki 305-8573, Japan E-mail: {hana, kitagawa}@kde.is.tsukuba.ac.jp Abstract Dissemination and exchange of a large amount of documents have become popular according to the advance of network technology in recent years. Thus, importance of content analysis techniques is increasing. Topic analysis in a series of large-scale document streams such as E-mail and news articles is one of such. important research issues. This research especially aims at the analysis of time varying activation levels of topics. When documents of high relevance with a specific topic arrive very frequently, then the activation level of the topic is regarded high, otherwise the activation level is considered to be low. In this research, we propose a systematic topic analysis method for document streams incorporating both document arrival rate and document relevance. Moreover, we evaluate the effectiveness of the proposed method by experiments using real data. Keyword document stream , topic analysis, relevance, document arrival rate