An improved incremental Naïve Bayesian for sentiment classification

作者：Yang Fengqin; Zhou Wanting; Wu Di^*; Sun Xiaoxin; Sun Hongguang; Sun Tieli

来源：ICIC Express Letters, 2013, 7(3 B): 959-964.

• Bayesian
• Effect of noise
• Movie reviews
• Sentiment analysis
• Sentiment classification
• Supervised methods
• Training dataset

Web review mining and sentiment analysis can help customers effectively. Currently the existing methods are mainly divided into supervised and unsupervised. This paper adopted supervised method and proposed an improved incremental Nai¨ve Bayesian (Named T-INB) for Sentiment Classification. In order to decrease the effect of noise data in traditional incremental Bayesian, a threshold is introduced and it can control the order in which data is added to the original training dataset. Experimental results on a movie review data set show our approach is feasible and effective.

• 单位
  东北师范大学