分析網頁文章以辨識讀者情感有助於彙整顧客評論、遞送網路廣告以及預測銷售與經濟趨勢。許多研究人員都致力於研究觀感分類,將網路上的非結構化文章分類到正向或負向觀感裡。然而,如何依據讀者的情感將文件做分類,依然缺乏相關的研究。本研究基於心理學所定義的基本情緒與心情開發情感分類器,以辨識網頁文章所能觸動的讀者情感。採用基本情感可以減少情感分類時的複雜度並且提供標準的情感類別。實驗結果顯示,本研究所開發的分類器可達到準確的分類效果;支持向量機分類器的分類效果優於單純貝氏與序列最小優化分類器。此研究結果可以用來改善各項電子商務應用,例如,廣告遞送系統、企業智慧系統、即時通訊以及線上聊天室。
Recognizing affects from Web articles is important for analyzing customer reviews, delivering ads, and predicting sales and economic trends. Many researchers have devoted themselves to studying sentiment classification in order to classify unstructured texts on the Web as having positive or negative sentiments. However, few of them addressed how to classify documents on the basis of readers' affects. This study developed affect classifiers based on basic emotions and moods as defined in psychology, instead of subjective emotion/mood categories, to decrease the ambiguity and confusion. News articles were collected from the Web and labeled with basic emotion and mood categories for training the classifiers. The experimental result showed that this approach can achieve good performance and that SVM classifiers are more effective than naïve Bayes or SMO classifiers. The electronic-commerce applications such as online ad delivery systems, business intelligence systems, instant messengers and online chat rooms can be designed based on the proposed approach.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。