This paper proposes a refined Hierarchical Dirichlet Process (HDP) model for unsupervised Chinese word segmentation. This model gives a better estimation of the base measure in HDP by using a dictionary-based model. We also show that the initial segmentation state for HDP model plays a very important role in model performance. A better initial segmentation can lead to a better performance. We test our model on PKU and MSRA datasets provided by Second Segmentation Bake-off (SIGHAN 2005) [1] and our model outperforms the state-of-the-art systems. © Springer-Verlag 2013.
CITATION STYLE
Pei, W., Han, D., & Chang, B. (2013). A refined HDP-based model for unsupervised Chinese word segmentation. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 8202 LNAI, pp. 44–51). https://doi.org/10.1007/978-3-642-41491-6_5
Mendeley helps you to discover research relevant for your work.