[目的/意義]粗粒度的跨語言情感分析往往只考慮篇章段落中的情感分布信息,而忽略深層次的語義關系。為彌補這一缺陷,文章提出一種基于依存句法的跨語言細粒度情感分析算法。[方法/過程]首先對中文文本進行依存句法分析,從深層的語義關系入手,提出一系列識別規則,提取情感評價單元;然后利用機器翻譯對情感評價單元進行語言轉換;最后通過與英文情感詞典的匹配分析,對情感評價單元的情感極性進行判定。[結果/結論]實驗結果表明,與原有的情感評價單元提取方法相比,本文的方法在一定程度上提高了提取效率;同基礎的單語言情感分析方法相較,借助英文情感詞典所做的跨語言情感分類結果更為理想。

[Purpose/significance] The document-level cross-lingual sentiment analysis usually focuses only on the distribution of emotional information in the whole document,but hardly takes the deep semantic relations into account. The fine-grained crosslingual sentiment analysis algorithm based on dependency parsing introduced in this paper can be used to overcome this disadvantage. [Method/process] This paper first parses the Chinese document by the means of dependency parsing. Starting with the deep semantic relation, the paper puts forward a series of identification rules and extracts sentiment appraisal units based on the rules. And then the paper translates the Chinese sentiment appraisal units into English with the aid of machine translation. Finally,by matching with the English sentiment dictionary,the paper identifies the sentiment polarity of the sentiment appraisal units.[Result/conclusion] The experimental results show that compared with the original method of extracting sentiment appraisal units,the proposed method improves the extraction efficiency to some extent,and the results of the cross-lingual sentiment classification with the aid of English dictionary are proved better than the fundamental monolingual sentiment classification method.