透過您的圖書館登入
IP:3.140.188.16
  • 學位論文

SentiOntology:一個情感本體推論系統

SentiOntology: An ontology-based sentiment referring system

指導教授 : 洪智力
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


隨著網際網路的發展,評論文章的資料量龐大,在情感分析研究的領域中,如何將相關的電子文本有效且正確的分類情感傾向,情感自動化辨識與分類成為熱門研究議題之一。針對評論文章根據詞彙辨別,若能準確將情感詞彙建置進本體中,將會使機器達到自動化情感辨識。情感分析研究中,SentiWordNet為一個重要的情感詞彙資源,以WordNet為基礎將詞彙給予相對的情感分數來表示詞彙情感的程度,根據每個詞彙給予正面、負面、中立三個情感傾向數值,透過詞彙情感分數便能更加瞭解詞彙具有的情感。雖然給予相關的情感分數,在電影評論中不同類別出現相同詞彙,例如「恐怖」這個詞彙,在恐怖片與喜劇片中情感分數會不同,而這樣類型的詞彙在SentiWordNet情感分數中還是會有所差異。由於學者大多探討SentiWordNet對情感文章的影響,並沒有將情感自動化辨識,讓機器可以達到自動化分析情感。因此,本研究提出將結合SentiWordNet情感字典內負面詞彙,從SentiWordNet情感字典中萃取情感數值和有效的詞彙,並修正在不同類別詞彙分數的差異,建置詞彙概念至知識本體,達到自動推論的情感詞彙本體,詞彙經過本體推論後產生新的情感數值,應用SVM分類的結果顯示,經由本體推論後產生新的詞彙情感數值的分類結果高於SentiWordNet語料庫數值分類的結果。

並列摘要


With the rapid growth of available subjective text on the internet, the sentiment classification concerns the use of automatic methods for predicting the orientation of subjective content on electronic documents lead to an important opinion mining research issues. Based on the quantitative analysis of WordNet database where each term is associated with numerical scores indicating positive and negative sentiment information, SentiWordNet has become a public and popular lexical resource. Although SentiWordNet is useful to give words sentiment values and tendencies, it suffers from the feature of general purpose. For example, the sentiment score of the word of "Terror" is different between comedy and horror films. The main reason is different linguistic levels (words, sentences and documents) rely on corpus-based and lexicon-based methods in sentiment classification. Therefore, this study developed an Ontology-Based sentiment inference method which synthesized the positive, negative words extacted from SentiWordNet sentiment dictionary, and corrected the vocabulary scores errors in the different categories by revised sentiment value and tendency for objective words in SentiWordNet based on assessment of the co-relevance of each objective word and its associated sentiment sentences to sentiment classification automatically.

參考文獻


王誌鴻(2007)。發展知識本體機制於中文新聞解析與分類(碩士論文)。取自華藝線上圖書館。(系統編號U0017-1110200712095971)。
戚玉樑、蔡明宏(2006)。以文件為對象的概念萃取程序建立知識本體的雛型架構。資訊管理學報,14,47-66。
陳玉鳳(2008)。發展以語意規則為基礎的新聞分派系統(碩士論文)。取自華藝線上圖書館。(系統編號U0017-0707200915423504)。
Abulaish, M., Jarhiruddin, Doja, M. N., & Ahmad, T. (2009). Feature and opinion mining for customer review summarization. Lecture Notes in Computer Science, 5909, 219-224.
Berners, L. T., Hendler, J., & Lassila, O. (2001). The semantic web. Scientific American, 284(5), 36-43.

延伸閱讀