面向在线新闻的公众情感计算与舆情分析关键技术

Social emotion detection and public opinion analysis oriented online news provide decision makers with potential opportunities and convenience of effective public management, through timely and accurate understanding of public opinions and attitudes. Nevertheless, there are many practical challenges in social sentiment analysis and public opinion analysis. Such as, public sentiment dictionary can't reflect the context information, emotional detection models ignore emotion dependency, the accuracy and adaptability of the new event detection model will be improved and so on. The project will focus on the following issues: (1) We propose an efficient algorithm to automatically build an emotional dictionary context-based for social emotion detection. (2) Investigate the problem of emotion dependency and propose an effective emotion dependency model to improve emotion detection accuracy. We demonstrate that social emotions are not independent but are correlated with each other by their homogeneous or heterogeneous nature. The dependencies between emotions can reflect more emotional connections of online news articles and therefore are beneficial to the detection of social emotions. (3) A new term weighting scheme for New Event Detection(NED) is proposed. The scheme can capture the uniqueness of each story; meanwhile, it can smooth the disturbing effect of synonymous and polysemous issues. The research results of this project should be the foundation of Social emotion detection and public opinion analysis oriented such social media as blogs, BBS, weibo/twitter and so on.

对在线新闻唤起的公众情感进行检测，及时把握当前舆情动态及正确引导网络舆情发展具有重要的现实意义。当前，面向在线新闻的公众情感计算和舆情分析仍然存在着很多问题，如公众情感词典不能反映上下文信息、情感检测模型缺少情感关联关系、新事件检测模型的精度和适应性有待提高等。因此，本项目针对面向在线新闻的公众情感计算与舆情分析的准确性和高效性的需求，重点研究：(1) 研究构建基于上下文的情感词典，提高基于情感词典的公众情感计算的准确度；(2) 研究构建基于情感关联关系的公众情感检测模型，使得用于预测事件可能产生的公众情感，从而可以掌握相关事件的发展动态；(3) 研究构建一种新的词加权方法用于新事件检测，使得不仅可以表达每篇新闻报道的唯一特性，也可以减少一词多义和多词同义对新事件检测带来的影响。本项目的研究成果也将成为研究其他社交媒体（如博客、论坛、微博等）的公众情感与舆情分析的基础。

对在线新闻唤起的公众情感进行检测，及时把握当前舆情动态及正确引导网络舆情发展具有重要的现实意义。当前，面向在线新闻的公众情感计算和舆情分析仍然存在着很多问题，如公众情感词典不能反映上下文信息、情感检测模型缺少情感关联关系、新事件检测模型的精度和适应性有待提高等。因此，本项目针对面向在线新闻的公众情感计算与舆情分析的准确性和高效性的需求，重点研究了：(1) 研究构建基于上下文的情感词典，提高基于情感词典的公众情感计算的准确度；(2) 研究构建基于情感关联关系的公众情感检测模型，使得用于预测事件可能产生的公众情感，从而可以掌握相关事件的发展动态；(3) 研究构建一种新的词加权方法用于新事件检测，使得不仅可以表达每篇新闻报道的唯一特性，也可以减少一词多义和多词同义对新事件检测带来的影响。本项目的研究成果也将成为研究其他社交媒体（如博客、论坛、微博等）的公众情感与舆情分析的基础。