A Short Text Classification Method Based on Supervised Biterm Topic Model
In response to the problems of semantic sparsity and ambiguity in short texts,this paper proposes a Supervised Biterm Topic Model(Su-BTM)and applies it to short text classification.Based on the BTM topic model,distribution parameter between topic and category is introduced to identify semantic information between topic and category,accurate mapping between topic and category is established,and a Su-BTM-Gibbs topic sampling method is proposed to sample the implied topics of each word.Comparative experiments are conducted on two datasets of Chinese and English short texts,and the results show that this method has better classification performance compared to classical models.
semantic sparsityBTM topic modelimplied topicshort text classification
国家科技期刊平台
NSTL
万方数据
