基于监督学习的微博情感分类方法

SUPERVISED LEARNING-BASED MICROBLOGGING SENTIMENT CLASSIFICATION METHOD

随着在线社交网络的快速发展,微博平台上聚集了大量的包含情感的主观句。微博情感可影响受众的观点形成,作用于商务智能、政策制定,甚至是股票市场。微博情感分类是指如何从微博中自动抽取出情感极性和不同的情感分类,如喜爱、愤怒、惊奇等。结合情感词汇本体和同义词词林,从微博中抽取不同类别的特征,运用监督学习方法进行情感分类,在学习过程中优化不同的模型,并分别进行误差和拟合分析,比较不同模型的性能。分类算法在NLP&CC 2013的评测任务中取得了具有竞争性的结果。

With the rapid development of online social networks,microblogging platforms gather a lot of subjective sentences which contains the sentiment. The sentiment in a microblog can affect the formation of audiences＇ opinion,and can act on business intelligence,policy development,and even the stock market. Microblogging sentiment classification refers to how to automatically extract the emotion polarity and different emotional categories,such as love,anger,surprise,etc,from microblogs. Combined with emotional lexicon ontology and synonymy thesaurus,we extracted the characteristics with different categories from microblogging text,and used the supervised learning method to classify the microblogging sentiment. Different models are optimised in learning process. The error analysis and fitting analysis are processed respectively as well,and the performances of different models are compared. This classification algorithm achieved competitive result in NLP CC 2013 evaluation task.