本科毕业论文-微博文本情感分析研究与实现.docxVIP

微博文本情感分析研究与实现 摘 要 随着互联网的发展，越来越多的人使用在线社交网站来表达自己的情感或分享自己对某件事情的看法，也因此互联网上的数据呈现爆炸式增长。分析大量数据背后所包含的情感信息，有助于人们及时了解社会上针对某一产品、话题等的舆论态度，进而可以对负面的情感和态度做出及时的预警，提醒相关人员或组织及时采取应对措施。 文本情感分析是自然语言处理领域的一个重要的研究方向，目前主流的情感分析方法是抽取文本特征并构建有监督的机器学习模型。抽取的文本特征与所采用语料的领域、语法特点等有很大的关系，因而往往需要针对特定的语料选择相适应的文本特征。而在机器学习的模型选择方面，也需要根据训练集的特点和分析需求选择合适的分类或回归模型。 本文基于LDA主题模型和支持向量回归模型，针对新浪微博数据提出了一个文本情感分析算法。该算法结合微博数据短文本的特点，充分考虑文本所包含的语义，抛弃了常用的基于文本结构或统计信息的文本特征，而通过LDA模型将文本抽象为主题相关的文本向量。最后使用产生的文本向量训练支持向量回归模型，并对微博数据的情感倾向和情感强烈程度进行标注。实验表明，该算法在对于微博数据的情感分析上有一定的效果。 最后，文本为了便于人们对标注了情感的数据进行直观的分析，提出并实现了一个可交互的可视化应用系统。人们可以使用此系统对某一话题或某一用户的情感大体情况和走势有一个直观的了解。 关键词：舆情监测，情感分析，文本分类，数据可视化  Research and Implementation of Sentiment Analysis on Weibo Data Abstract With the development of the Internet, an increasing number of people share their emotion statuses or attitudes on online social websites, leading to an explosive rise on the scale of data on the Internet. Mining sentiment information behind data helps people know about public opinions to a product or topic. By doing this, negative attitude and emotion can also be detected so that organizations or governments can take measures in time. Text sentiment analysis is a basic research area in natural language processing. At present most of sentiment analyzing algorithms build supervised machine learning models by extracting text features from texts. Text features is much relative to the area, structure feature of texts, so it is necessary to select features considering the characteristics of the corpus. When selecting machine learning models, it is also important to use an adapting classification or regression model. In this paper a sentiment analysis algorithm adapting to Weibo data is proposed based on LDA topic model and SVR machine learning method. Given that a Weibo post is usually short, text features based on structure or statistics are deprecated. And taking topics and semantic information behind texts into consideration, LDA model is used to generate text features. To decide