Disambiguating Dynamic Sentiment Ambiguous Adjectives

Proceedings of the 23rd International Conference on Computational Linguistics (Coling 2010), pages 1191–1199, Beijing, August 2010 Disambiguating Dynamic Sentiment Ambiguous Adjectives Yunfang Wu �Key Laboratory of Computational ��Linguistics (Peking University), Ministry of Education��� China wuyf@pku.edu.cn Miaomiao Wen Department of Electrical Engineering and Information Systems, University of Tokyo wenmiaomiao98@gmail.com Most of the work was performed when the author was a student at Peking University. Abstract Dynamic sentiment ambiguous adjectives (DSAAs) like “large, small, high, low” pose a challenging task on sentiment analysis. This paper proposes a knowledge-based method to automatically determine the semantic orientation of DSAAs within context. The task is reduced to sentiment classification of target nouns, which we refer to sentiment expectation instead of semantic orientation widely used in previous researches. We mine the Web using lexico-syntactic patterns to infer sentiment expectation of nouns, and then exploit character-sentiment model reduce noises caused by the Web data.to At sentence level, our method achieves promising result with an f-score of 78.52% that is substantially better than baselines. At document level, our method outperforms previous work in sentiment classification of product reviews. 1 Introduction In recent years, sentiment analysis has attracted considerable attention in the NLP community. It is the task of mining positive and negative opinions from natural language, which can be applied to many research fields. Previous work on this problem falls into three groups: opinion mining of documents, sentiment classification of sentences and polarity prediction of words. Sentiment analysis both at document and sentence level rely heavily on word level. The most frequently explored task at the word level is to determine the polarity of words, in which most work centers on assigning a prior polarity to words or word senses in the lexicon out of context. However, for some words, the polarity varies strongly with context, making it hard to attach each to a fixed sentiment category in the lexicon. For example, the word “low” has a positive orientation in “low cost” but a negative orientation in “low salary”. We call these words like “low” dynamic sentiment ambiguous adjectives (DSAAs). Turney and Littman (2003) claim that DSAAs cannot be avoided in a real-world application. But unfortunately, DSAAs are discarded by most research concerning sentiment analysis. In this paper, we are devoted to the challenging task of disambiguating DSAAs. The task is to automatically determine the semantic orientation (SO) of DSAAs within context. We limit our work to 14 frequently used adjectives in Chinese, such as “large, small, many, few, high, low”, which all have the meaning of measurement. Although the number of such ambiguous adjectives is not large, they are frequently used in real text, especially in the texts expressing opinions and emotions. As demonstrated by the experimental results in this paper, the disambiguation of 14 DSAAs can obviously improve the performance of sentiment classification of product reviews. The task of disambiguating DSAAs is reduced to sentiment classification of nouns. Previous studies classify nouns into three categories: positive, negative and neutral. In contrast, we propose two categories of …

Content truncated to 500 words