陳建錦Chen, Chien Chin臺灣大學:資訊管理學研究所曾有德Tseng, You-DeYou-DeTseng2010-05-052018-06-292010-05-052018-06-292009U0001-2207200911552800http://ntur.lib.ntu.edu.tw//handle/246246/180016Web2.0的盛行使得網際網路成為重要的商業資訊來源，透過許多電子商務網站所提供的商品評論平台，網際網路使用者可自由地撰寫商品相關的評論，正面的商品評論可幫助消費者制定商品的購買決策，而負面的商品評論可協助企業檢討與修正商品的商業策略。但隨著評論數量快速地增長，消費者與企業均需要有效的資料探勘技術來由大量的文字資訊中找出重要的評論意見。現行的評論意見探勘技術多忽略了評論內容的資訊品質，以致於探勘出的評論其資訊品質良莠不齊。在本研究中，我們提出一套方法來評估商品評論的資訊品質，我們將資訊品質評估視為一種分類問題，並使用一套有效的資訊品質架構來萃取重要的評論資訊特徵。實驗結果顯示我們提出的方法有優異的資訊品質評估效能，而且顯著地優於其它學者在近幾年所提出的方法。此外本研究還進行升力曲線分析找出高品質評論所具備的重要因素。最後我們提出一個以評論品質分類器為基礎的評論檢索雛型系統，來幫助使用者有效地搜尋到包含他們需要的有用資訊之評論。The ubiquity of Web 2.0 makes the Internet an invaluable source of business information. For instance, product reviews composed collaboratively by many independent Internet reviewers can help consumers make purchase decisions and enable enterprises to improve their business strategies. As the number of reviews is increasing exponentially, opinion mining is needed to identify important reviews and opinions to answer users’ queries. Most opinion mining approaches try to extract sentimental or bipolar expressions from a large volume of reviews. However, the mining process often ignores the quality of each review and may retrieve useless or even noisy documents. In this thesis, we propose a method for evaluating the quality of information in product reviews. We treat the evaluation of review quality as a classification problem and employ an effective information quality framework to extract representative review features. Experiments based on an expert-composed data corpus demonstrate that the proposed method outperforms state-of-the-art approaches significantly. Moreover, this thesis implements detailed lift analyses to find the important factors for constructing high-quality reviews. Finally, we propose a prototype of review retrieval system that based on the classifier of review quality to help users to efficiently search the reviews that contain helpful information they want.論文摘要 iHESIS ABSTRACT iiable of Contents iiiist of Figures vist of Tables vihapter 1 Introduction 1.1 Background 1.2 Motivation 2.3 Objectives, Approaches, and Results 4.4 Thesis Organization 5hapter 2 Related Works 6.1 Opinion mining 6.1.1 Opinion Extraction and Polarity Identification 6.1.2 Opinion Target Identification 7.2 Opinion Retrieval 8.3 Review Quality Evaluation 9.3.1 Ranking-based methods 10.3.2 Classification-based method 11hapter 3 Methods 13.1 Definition of Review Quality 13.2 Information Quality-based Review Features 15.3 Classification Models 19.3.1 The Binary SVM 20.3.2 The Kernels 23.3.3 Multiclass SVMs 25hapter 4 Performance Evaluations 27.1 Data Preprocessing and Annotation 27.1.1 Description of Data Annotation 27.1.2 Evaluating the Agreement of Annotations 28.1.3 Experiments Design 31.1.4 Metrics for Evaluating Performance 31.2 IQ Dimension Evaluations 34.3 Comparisons with Other Methods 37.4 High-Quality Review Analysis 39hapter 5 Quality-based Review Retrieval System 44.1 Data Preprocessing 45.2 Classification Model Construction 46.3 Review Quality Evaluation 46.4 Review Ranking and Retrieval 46hapter 6 Conclusions 49.1 Contributions 49.2 Future Works 50eferences 52ppendix 1. The Definitions of Wang and Strong’s IQ Framework [27] 55application/pdf998059 bytesapplication/pdfen-US文件探勘分類意見探勘資訊品質商品評論支撐向量機評論檢索系統text miningclassificationopinion mininginformation qualityproduct reviewssupport vector machinereview retrieval systemhttp://ntur.lib.ntu.edu.tw/bitstream/246246/180016/1/ntu-98-R96725044-1.pdf