【摘要】 研究结合社交媒体特点,充分考虑标签文本和内容文本信息,融合了传统的LDA话题模型对社交文本信息进行话题聚类,从而实现了对社交数据的话题发现,与此同时,文章提出了基于关键词图模型构建话题特征,并结合支持向量机模型进行文本情感极性判别。研究在开放微博数据集和COAE2014公开评测数据上进行了相关实验,实验证明了有效的关键词图模型能进一步克服中文语义的模糊性和歧义性。更多还原

【Abstract】 Social media is the platform for our daily digital life,social topic detection is a hot but difficult issue for data in social media is with the complexity in heterogeneity,timing and linguistic ambiguity. In this paper,we apply Maximum Entropy on tag texts and LDA model on social contents for social topic detection,meanwhile,a key word graph based method is proposed for text sentiment analysis with SVM. Experiments on open Weibo data and COAE2014 data show the effectiveness of our proposed strategy for Chinese semantic analysis.更多还原