基于主谓情感差异性句法分析框架的跨语言情感分析

Sub-pre Distinct Emotional Differences Analysis Framework for Cross-lingual Sentiment Analysis

随着互联网的快速发展和全球化进程的加快,因特网提供的信息资源呈现出多语言化的特点.各语言可利用资源的不平衡给网络信息的研究和利用带来障碍.在此背景之下,利用资源丰富语言对资源贫乏语言进行情感分析具有重要的应用价值与现实意义.提出一种基于主谓情感差异句法分析框架,实现跨语言的情感分析.将目标语言数据翻译后作为测试集,利用句法分析工具Stanford Parser进行句法分析,结合情感词典和转折词表,再基于双语语料训练出贝叶斯分类器后对跨语言测试语料进行最终的情感分析.实验结果表明,本文方法在情感分类的准确率上高于传统方法 20%以上,具有一定的实用价值.

With the rapid development of the Internet and the acceleration of the globalization process,available information resources on the Internet showmulti-language-oriented features. The imbalance of available resources in different language hinders the research and utilization of network information. Based on this background,using resourceful language（ as source language） to help the less resourceful language＇s（ as target language） sentiment analysis has important application value and practical significance. In this paper,we present a framework of cross-language sentiment analysis,which is based on subject-predicate distinct syntactic analysis. It treats the translated target language data as the train set and uses the Stanford Parser to parse sentences. Combined with the emotional dictionary and transition words,we train the Bayesian classifier for the final sentiment analysis of the cross-language test set. The experimental results showthat our method achieves better sentiment classification accuracy 20% higher than the traditional method,and has certain practical value.