CRF Based on LHFS Applied on Sentiment Classification

Sentiment classification has attracted increasing interest from natural language processing. This paper applies CRF(Conditional Random Filed) based on LHFS (Local High-Frequency Strings) method on sentence sentiment analysis. This method can effectively solve ordinal regression problems. In this method, sentences are labeled to determine their polarity, and LHFS method is used to expand the set of sentiment features. Experiments on sentiment classification indicate that the accuracy of CRF model is increased up to 2.1%, with the help of LHFS method, which is much better than that of HMM and MEMM.