Affiliation:
1. Communication University of China
2. Beijing University of Posts and Telecommunications
Abstract
There exist principles of Relevancy, Orthogonality and Domain Adaptation for auxiliary problems (APs) selection in Alternating Structure Optimization (ASO) algorithm. Convex Alternating Structure Optimization (cASO) algorithm is an improved one based on ASO algorithm, whose kernel still lies in creating excellent APs. In order to validate the effectiveness of the preceding principles in cASO algorithm, many types of APs were created by taking example of Chinese syntactic chunking. Experimental results and analyses both demonstrate that those principles still hold in cASO algorithm.
Publisher
Trans Tech Publications, Ltd.
Reference15 articles.
1. R.K. Ando and T. Zhang. Journal of machine learning research, 6(Nov), 2005, 1817-1853.
2. R.K. Ando and T. Zhang: A High-performance Semi-supervised Learning Method for Text Chunking. Proc. 43rd annual meeting on association for computational linguistics, Michigan, 2005, 1-9.
3. X. Bai, T. Zhang, S. He and X. Wang: Chinese Syntactic Chunking Based on ASO Algorithm (In Chinese). Proc. 13th China national conference on artificial intelligence (CAAI-13), Beijing, (2009).
4. C. Liu and H.T. Ng: Learning Predictive Structure for Semantic Role Labeling of Nombank. Proc. 45th annual meeting on association for computational linguistics, Prague, 2007, 208-215.
5. S. He, T. Zhang, X. Bai, et al: Incorporating Multi-task Learning in Conditional Random Fields for Chunking in Semantic Role Labeling. Proc. IEEE international conference on natural language processing and knowledge engineering (IEEE NLP-KE'09), Dalian, 2009, 47-51.