Abstract: With the rapid development of the Internet，more and more opinionated texts are surging on the web.It is of practical and commercial value to mine opinion words and classify their polarities in these texts.A translation approach is proposed to extract opinion words，using Chinese-English Machine Translation（MT） system to convert the Chinese seed opinion words into English words.The WordNet is employed to get hypernyms，hyponyms，synonyms of these English words and use English-Chinese MT system to translate them into corresponding Chinese words which are used as candidate opinion words.Additionally，the SentiWordNet is used to classify English opinionated word polarities，which are to be mapped to those of the Chinese opinion words.Experimental results show that the approaches taken in this paper improve the precision of mining opinionated words and classifying their polarities.

Key words: opinionated words, polarity classification, translation approach

摘要： 随着互联网的扩展，网络上出现了越来越多的含有观点信息的主观性评论文本。挖掘这些文本中的情感词语并进行极性判别具有重要的现实意义和商业价值。为此，提出一种基于翻译方法的情感词提取方法，使用汉英机器翻译系统翻译汉语种子情感词典生成候选英语词语，根据WordNet提取候选英语词语的上下位词、同义词或反义词并将这些词语翻译成汉语，进而提取汉语情感词语。另外，依据SentiWordNet判别候选英语词语极性，并将候选英语词语极性映射到目标汉语情感词语上，进而达到判别汉语情感词语极性的目的。实验结果表明上述方法可以有效提高情感词的识别效率以及极性判别的准确率。

关键词: 情感词语, 极性判别, 翻译方法