Apply Document Processing Techniques to Improve Fuzzy Formal Concept Analysis Concept Quality

• Main Authors: Jin-De Peng, 彭晉德
• Other Authors: Chuen-min Huang
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/13267447827823741877

碩士 === 國立雲林科技大學 === 資訊管理系碩士班 === 101 === Traditional Formal Concept Analysis (FCA) has been blamed for its drawback that fails to deal with uncertain information; furthermore, its performance to administer and searching deteriorated while coping with a large amount of documents and particularly in wider domains. To deal with this drawback, this study employed Fuzzy Theory into FCA and used Event Detection Clustering Technique based on the experimental dataset from Yahoo! News. We extracted news features through syntax rules and assigned normalized TF-IDF as membership grade. Then, event detection clustering was carried out to decrease the complexity of document set, enhance quality of searching and shorten the duration of processing. In comparison with traditional FCA, the results showed that our proposed method, assessing the quality of concept lattice through fuzzy rate, had higher fuzzy rate and the quality also increased as α-cut was higher. Furthermore, we are able to find an appropriate α-cut to build more precise concept lattice through users’ satisfaction. Experimental results showed that users indicate that the concept expressed the news contents best as the α-cut was 0.06 in this study.