Perceptron-based tagging of query boundaries for Chinese query segmentation

Jingfei Du, Yan Song, Chi-Ho Li

Abstract: Query boundaries carry useful information for query segmentation, especially when analyzing queries in a language with no space, e.g., Chinese. This paper presents our research on Chinese query segmentation via averaged perceptron to model query boundaries through an L-R tagging scheme on a large amount of unlabeled queries. Experimental results indicate that query boundaries are very informative and they significantly improve supervised Chinese query segmentation when labeled training data is very limited.