A CharacterBased Joint Model for CIPSSIGHAN Word Segmentation Bakeoff 2010
Kun Wang, Chengqing Zong, Keh‐Yih Su · 2013
This paper presents a Chinese Word Segmentation system for the closed track of CIPS-SIGHAN Word Segmentation Bakeoff 2010. This system adopts a character-based joint approach, which combines a character-based generative model and a character-based discriminative model. To further improve the crossdomain performance, we use an additional semi-supervised learning procedure to incorporate the unlabeled corpus. The final performance on the closed track for the simplified-character text shows that our system achieves comparable results with other state-of-the-art systems. 1