- 學位論文
以評價理論為基礎的電子本文分類
The Document Classification Based On The Appraisal Theory
摘要
網路口碑具有非常大的行銷價值，也是企業與產品在網路上的信用資產，透過網路口碑的搜尋，企業可以得知自己產品的市場反應，例如：產品品質、服務品質以及顧客抱怨，根據這些網路口碑背後的隱藏資訊，往往能夠協助企業進行行銷策略的擬定，拉近企業與客戶之間的距離，進一步探討這些評價，可發現使用者對於評價的情緒反應，更直接影響企業商譽。文章情感分析的技術可以協助企業找出口碑背後的隱藏價值，傳統的文章情感分析多半藉由演算法與統計的技術來區分文章情感面向，例如：常見的演算法包含SVM (Support Vector Machines)、貝式分類，統計方式則有Entropy與馬可夫鏈等等的方式。也有其他學者直接以人工方式建立控制字彙來區分情感面向，本研究以評價理論架構，應用在情感分析的建構上，透過客觀的角度建立控制字彙，將級差系統中的情感值強弱觀念導入，除了能將文章進行情感分類與與情感分級，並與其他情感分析的做法進行比較，結果發現評價理論字彙在分類上的準確度與穩定性，比其他的方式來的優異。
並列摘要
In recent years, Internet word-of-mouth marketing has been more and more important for the enterprise. Many enterprises obtain the information of the products by internet word-of-mouth, such as the product quality, service quality and the customer complaints. We can develop marketing strategies to improve the customer relationship through these information. In order to further discuss and analyze the information, many enterprise began to think highly of the technology of the opinion mining, and try to explore the opinions form the users . Further discuss these opinion, we find that the appraisal behind the user emotional express the enterprise image. The sentiment analysis aims to mind customers’ thoughts through mining the sentimental perspectives of those criticisms and comments. The algorithms and statistics was often used to solve the problems of the sentiment analysis, but this common methods were few theoretical foundation. In this study, we build the controlled vocabularies based on the appraisal theory as SVM features. The system of the appraisal theory was based on the linguistic resource, there are three sub-system that contain the attitude system, the engagement system and the graduation system. According to the graduation system, we establish the emotional feature which quantifies the emotion. In this experimental result, we confirm that the controlled vocabularies based on the appraisal theory can improve the accuracy and stability in SVM algorithm.