結合多辭典與常識網路的情緒分析系統

Sentiment Analysis Using Multi-dictionary and Commonsense Knowledgebase

吳蕙欣(Hui-Hsin Wu)

指導教授 ： 許永真

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2011年)

https://doi.org/10.6342/NTU.2011.02347

摘要

本研究為基於多種情緒辭典與常識網路以協助分析歌詞文本之情緒。此研究可應用於以情緒為主的推薦系統或搜尋引擎等相關研究。由於現今除了英語的情緒詞語資源較豐富外，其餘語言則常因為情緒詞語資源的不完善，不容易挖掘文本情緒或是作更進一步的應用。因此，提出一語言獨立的情緒詞語擴散方法來得到較完善的情緒詞語辭典是本研究的重點。

目前情緒分析的應用，往往只基於一種情緒辭典，我們為了增加的情緒詞語資料的完整性，收集了九種不同類型的情緒辭典，並利用辭典間相互驗證的方法來增加情緒詞語資料的正確性。

透過常識網路(ConceptNet)具有大量知識且概念與概念間互相連接的特性，藉由情緒擴散激發，將情緒詞語種子的情緒值擴散到相鄰的概念，傳遞到整個常識網路，得到擴散後的情緒辭典，我們稱之為iSentiDictionary。其包含28,248個詞語(9,701個單字與18,547的概念)，且每個詞語皆分配一個情緒分數，介於-1和1之間。

之後，我們利用所建構的iSentiDictionary預測歌詞文本情緒值，其情緒誤差距離為0.4568，比起利用翻譯ANEW情緒辭典的誤差距離0.7315，降低了0.2747。

關鍵字

歌詞情感 ； 情感分析 ； 意見挖掘 ； 常識網路 ； 情緒辭典

並列摘要

This thesis presents a new approach to language independent sentiment analysis that combines multi-dictionary and commonsense knowledgebase.

Sentiment analysis is the task of identifying positive and negative opinions, emotions, and evaluations. One major impediment to Non-English sentiment analysis research is the lack of a complete sentiment dictionary. In light of this, we collected nine kinds of sentiment dictionaries as sentiment concept seed, then through sentiment spreading activation from common sense network (ConceptNet) to get more sentiment concepts. And got a sentiment dictionary named iSentiDictionary. iSentiDictionary contains 28,248 sentiment terms (9,701 words and 18,547 concepts), and assigned a sentiment score between -1 and 1 for each sentiment term.

Final, we used iSentiDictionary to mine sentiment from Chinese pop song dataset (iPop).Compared to use the translation of ANEW as sentiment dictionary, iSentiDictionary reduced the error distance from 0.7315 to 0.4568.

並列關鍵字

lyric sentiment ； sentiment analysis ； opinion mining ； commonsense knowledgebase ； sentiment dictionary

參考文獻

[2] L. Balkwill and W. F. Thompson. A cross-cultural investigation of the perception of emotion in music: Psychophysical and cultural cues: Psychophysical and culturalcues. Music Perception, 17(1):43–46, 1999.

[5] E. Cambria, A. Hussain, C. Havasi, and C. Eckl. Sentic computing: Exploitation of common sense for the development of emotion-sensitive systems. Development of Multimodal Interfaces: Active Listening and Synchrony, pages 148–156, 2010.

[9] P. Dodds and C. Danforth. Measuring the happiness of large-scale written expression: Songs, blogs, and presidents. Journal of Happiness Studies, 11:441–456, 2010.

[10] A. Esuli and F. Sebastiani. Sentiwordnet: A publicly available lexical resource for opinion mining. In Proceedings of LREC, volume 6, pages 417–422. Citeseer, 2006.

[14] S. Kamvar and J. Harris. We feel fine and searching the emotional web. In Proceedings of the fourth ACM international conference on Web search and data mining, pages 117–126. ACM, 2011.