Chinese text sentiment orientation identification based on Chinese-characters

Existing sentiment identification technologies for Chinese texts are mostly based on Chinese words/phrases. However, we notice that Chinese characters are ideogram and many of them contain rich sentiment information. Hence, a Chinese text sentiment orientation identification model based on Chinese characters but not words or phrases is proposed in this paper. Contrasting to existent models, our model has many advantages such as no necessary of dictionaries, words segmentation and stop-words removal. Moreover, its dimension of feature vector is lower and its processing efficiency is higher. Experiments show that our model greatly simplifies the identification process and improves the computational efficiency at very little loss of accuracy.