Chinese Word Segmentation With Character Abstraction

Chinese word segmentation is an important and necessary problem to analyze Chinese texts. In this paper, we focus on the primary challenges in Chinese word segmentation: low accuracy of out-of-vocabulary word. To resolve this difficult problems, we group the "similar" characters to generate more abstract representation. Experimental results show that character abstraction yields a significant relative error reduction of 24.83% in average over the state-of-the-art baseline.