Theme Classification for Web Forum Content

• Main Authors: Jia-Luen Tsai, 蔡佳倫
• Other Authors: Chao-Chang Chiou
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/74029617879366880025

碩士 === 元智大學 === 資訊管理學系 === 98 === Web forum has enormous topical discussion or questions. After automatic theme classification, users can easier acquire the information for each theme, and also make use of the valuable information. The expectation of research method, adaptive weighting classification, is flexible and simple computation to theme classification. First, our research use word segmentation and selecting specific part of speech in order to obtain theme-related words. Second, we applied MI to extract keywords, and realize these keywords belong to which theme. Meanwhile, grouping keywords as keyword set (KS).Third, we calculate the importance of each KS in a document. Forth, applying Genetic Algorithm (GA) to optimize weights, we calculate the weighted score in each theme. And we take the theme which has highest score as the theme of the document. In experiment, we classified 11 phone-related themes. And the classification result shows the F-Measure of each theme are about 30% to 45%.