口碑對消費者來說已經成為重要的參考依據。他們在做購買決策之前,會考慮其他消費者的經驗與建議。但是現在網際網路中有著大量的口碑資訊,加上評論者撰寫口碑文章的方式是沒有任何限制,導致消費者很難從眾多品質不一的文章中找到有用的資訊。在文獻中,大部分的口碑探勘研究主要都重視在文章情感的分類,卻忽略了口碑品質的重要性。本研究以資訊品質(IQ)架構為基礎來分類口碑文章,其中情感判斷方面,本研究採用SentiWordNet提供的情感值來判斷情感詞彙的傾向,進而自動化的辨識出文章中之意見句子。除此之外,不同的資訊品質因素在品質上可能有不同的影響力。因此,本研究萃取專家意見並透過層級分析程序法(AHP)來評估資訊品質因素的重要程度,再使用重要程度值加權資訊品質因素,使得資訊品質因素特徵值之間的差異更大。實驗結果顯示,本研究提出層級分析程序法技術整合資訊品質架構的方法能有效提升口碑品質分類的效果,也證實資訊品質當中不同的資訊品質因素具有不同的重要性。
WOM (word-of-mouth) has become important reference sources for consumers. They consider others’ experiences and suggestions before making purchase decisions. There are a lot of WOM be written by reviewers on the Internet. These reviewers’ writing styles are without any restrictions. For this reason, lead to difficult for consumers to find useful and qualitative information. In literature, most WOM mining researches mainly focus on document sentimental classification but ignore the importance of WOM quality. In this research, we classify WOM documents based on the information quality (IQ) framework. In the part of sentiment, we determine polarity of sentimental term with sentimental values of SentiWordNet in order to automatically identify the opinion sentences in a document. Furthermore, different elements of IQ may have different influencing power on quality. Therefore, this research extracts experts’ opinions and evaluates importance degree of IQ elements through the analytic hierarchy process (AHP). Experiment results show that our proposed approach which integrates IQ framework with AHP technique can promote performance on WOM quality classification and also demonstrate that different IQ elements have different importance on information quality.