隨著強調使用者參與的Web 2.0之蓬勃發展，越來越多Blog、論壇興起，也越來越多使用者會在網路上發表自己的評論，像是電影評論或是產品評論。這些線上評論對於人們而言，也變得是個重要且有用的資訊來源。然而，網路上的評論可能會多如牛毛，為了讓使用者能更輕鬆地閱讀評論，這些評論必須要經過處理，例如把這些評論自動分成正面評論、負面評論兩類，或是進行意見探勘。 在評論分類及意見探勘的研究上，形容詞是相當重要的因子。評論內的形容詞通常可以顯示出該評論的正、負面傾向。因此，以往的研究常以形容詞的正、負向及其出現頻率作為評論分類的基礎。為了提升對評論資訊應用的效率，本研究希望能擴展形容詞的可用特性。由於資訊擷取領域的相關研究顯示，文字出現在文章中不同的段落位置對文章之重要性有不同的意義。我們因此思考，評論中的形容詞在文章中所出現的位置，是否對評論而言也有不同的重要性？ 為了了解相關問題，本研究利用建立使用者興趣檔的概念，建構出正、負面評論的Opinion清單，其中輔以字詞權重的計算來區別各Opinion的重要性，期能以此來探索形容詞所處評論中的位置是否會對評論分類有所影響，而實驗結果也證實，形容詞在文章中的位置，確實在評論分類上是個具有功效的因子。 With the rapid developing of Web 2.0, more and more blogs and forums appear, and more and more people post reviews on the Web. The online review is becoming a useful and important information resource for people. However, the number of reviews can be in hundreds. These reviews must be processed so that user may read them easily, e.g. classifying them automatically into two polarities (positive and negative), or doing opinion mining. Adjective is an important factor in review classification and opinion mining. The adjectives in a review can usually show the orientation of the review, either positive or negative. Therefore, the semantic orientation of adjectives and the term frequency of adjectives are the basis of review classification in such researches. In order to increase the performance of the review classification, this research has aimed to expand the useful attribute of adjectives. Some researches of Information Retrieval show that the words in different paragraphs have different meaning to the document, so this research considers whether the adjectives in different paragraphs have different meaning to the review. To examine this adjective characteristic, this research has applied user profiles to under manage the position of the adjective to uncover its influence on the accuracy of review classification. The experimental results show that the position of adjectives in reviews is an effective factor on review classification.
