在文件分類的領域中貝氏分類已被廣泛的使用。本研究針對網路評論提出了改良式貝氏分類器,目的是從大量且非結構化的評論中判斷其中評論的情緒分類。有別於傳統的貝氏分類器(NBC)在計算情緒特徵詞機率值的問題。本研究提出了結合啟發式規則與貝氏分類器Integrated Heuristic Rules and Naïve Bayes classifier (IHRNBC)提升在情緒傾向判定時的分類效果。本研究應用在中文論壇的3C科技與美容保養兩個領域。實驗結果顯示,本研究所提出的方法(IHRNBC)在分類準確率上優於現有的分類器,如傳統的貝氏分類 NBC與支持向量機(SVM)。
The Naïve Bayes classifier (NBC) has been often used in classifying documents. Accordingly, this study proposes an improved Naïve Bayes classifier in an attempt to conduct sentiment orientation classification on a mass volume of unstructured online comments. The classifier proposed in this study is different from the general NBC, whose disadvantage is that it cannot handle the feature of different classes in calculating probability. Therefore, this research proposes an Integrated Heuristic Rules and Naïve Bayes classifier (IHRNBC) to enhance the classification performance of sentiment orientation in opinion mining. The proposed classifier is applied for the contents taken from customer reviews of 3C and beauty products posted on online Chinese forums. Results show that the proposed classifier performs better than existing classifiers such as the general NBC and the Support Vector Machine (SVM).