Using Bilingual Knowledge and Ensemble Techniques for Unsupervised Chinese Sentiment Analysis

Proceedings of the 2008 Conference on Empirical Methods in Natural Language Processing, pages 553–561, Honolulu, October 2008. c⃝2008 Association for Computational Linguistics Using Bilingual Knowledge and Ensemble Techniques for Unsupervised Chinese Sentiment Analysis Xiaojun Wan Institute of Compute Science and Technology Peking University Beijing 100871, China wanxiaojun@icst.pku.edu.cn Abstract It is a challenging task to identify sentiment polarity of Chinese reviews because the re- sources for Chinese sentiment analysis are limited. Instead of leveraging only monolin- gual Chinese knowledge, this study proposes a novel approach to leverage reliable English resources to improve Chinese sentiment analysis. Rather than simply projecting Eng- lish resources onto Chinese resources, our ap- proach first translates Chinese reviews into English reviews by machine translation ser- vices, and then identifies the sentiment polar- ity of English reviews by directly leveraging English resources. Furthermore, our approach performs sentiment analysis for both Chinese reviews and English reviews, and then uses ensemble methods to combine the individual analysis results. Experimental results on a dataset of 886 Chinese product reviews dem- onstrate the effectiveness of the proposed ap- proach. The individual analysis of the translated English reviews outperforms the in- dividual analysis of the original Chinese re- views, and the combination of the individual analysis results further improves the perform- ance. 1 Introduction In recent years, sentiment analysis (including sub- jective/objective analysis, polarity identification, opinion extraction, etc.) has drawn much attention in the NLP field. In this study, the objective of sen- timent analysis is to annotate a given text for polar- ity orientation (positive/negative). Polarity orientation identification has many useful applica- tions, including opinion summarization (Ku et al., 2006) and sentiment retrieval (Eguchi and Lavrenko, 2006). To date, most of the research focuses on English and a variety of reliable English resources for sen- timent analysis are available, including polarity lexicon, contextual valence shifters, etc. However, the resources for other languages are limited. In particular, few reliable resources are available for Chinese sentiment analysis1 and it is not a trivial task to manually label reliable Chinese sentiment resources. Instead of using only the limited Chinese knowl- edge, this study aims to improve Chinese sentiment analysis by making full use of bilingual knowledge in an unsupervised way, including both Chinese resources and English resources. Generally speak- ing, there are two unsupervised scenarios for “bor- rowing” English resources for sentiment analysis in other languages: one is to generate resources in a new language by leveraging on the resources available in English via cross-lingual projections, and then perform sentiment analysis in the English language based on the generated resources, which has been investigated by Mihalcea et al. (2007); the other is to translate the texts in a new language into English texts, and then perform sentiment analysis in the English language, which has not yet been investigated. In this study, we first translate Chinese reviews into English reviews by using machine translation services, and then identify the sentiment polarity of English reviews by directly leveraging English resources. Furthermore, ensemble methods are employed to combine …

Content truncated to 500 words