Applying Feature Extraction Technique to Build Topic Hierarchy Model for News Topic Chronicle

• Main Authors: Leng-chin Chen, 陳泠謹
• Other Authors: Chuen-min Huang
• Format: Others
• Language: en_US
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/62559068987940080161

碩士 === 國立雲林科技大學 === 資訊管理系碩士班 === 95 === The news reports have become an important way to acquire knowledge. In order to solve multiple topics problem, this study proposed a feature extraction technique to automatically construct a topic hierarchy. The fundamental idea of this technique is to extract a set of common features from the event clusters for topics. In addition, this study proposed a novel presentation of news information which is named Topic Chronicle. We employed the multi-document summarization technique to generate Topic Chronicle and provided person-oriented and event-oriented aspects of topic hierarchy for readers that they could choose a suitable topic according to their need. The information quality experiment showed that the average F-measure of person-oriented topic is 0.78 and the average F-measure of event-oriented topic is 0.53. The system quality experiment suggested that more than 70% of participants agreed that the quality of Topic Chronicle System is satisfactory. Topic Chronicle could expound the connotation of topic development to help readers understand the overall scenario.