Integration of Topic Hierarchies without Mutually Labeled Data

• Main Authors: Chi-Wei Hung, 洪啟偉
• Other Authors: Jyh-Jong Tsay
• Format: Others
• Language: en_US
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/72828003476969356494

碩士 === 國立中正大學 === 資訊工程研究所 === 92 === In the problem of integrating documents from different sources into a comprehensive topic hierarchy, the objective is to develop efficient techniques that improve the accuracy of traditional categorization methods by incorporating categorization information provided by data sources into categorization process. Notice that in the World-Wide Web, categorization information is often available from information sources. Observe that many of the topic hierarchies adopted by current information sources are highly related. We believe that categorization information can be used to improve classification accuracy. However, this kind of problem need mutually labeled data between two hierarchies. Maybe we have no enough mutually labeled documents between two hierarchies in the World-Wide Web, or even no mutually labeled data completely. In the thesis, we study the problem of integrating documents from different sources into a comprehensive topic hierarchy without mutually labeled data. To solve this problem, the Bayesian Extension algorithm will need a predicting algorithm. We present several techniques that predict relations between topic hierarchies and incorporate categorization information from source hierarchies into traditional classification methods. Experiment on collections from Openfind and Yam, and Google and Yahoo, well-known popular web sites, shows that incorporating predicted mapping from source hierarchies to target hierarchies can improve the classification accuracy.