Title: A Knowledge-Based Co-Location Model for Natural Language Processing.
Speaker: Professor Wen-Lian Hsu, Academia Sinica

　　Co-location is an important feature in Natural Language Processing (NLP). The familiar bigram, trigram, or N-gram models can all be cast in this framework. In our prior work, we have considered meaningful word pairs (as well as sense pairs), such as noun-verb, noun-noun, adjective-noun, and adverb-verb pairs. We collected over two million such pairs in a semi-automatic fashion, which proved to be extremely valuable in our Chinese syllable-to-word conversion system. In this talk we shall consider various extensions of co-location relations, the methodology to collect them, and how they can be combined together to form a knowledge base that could be useful for an array of important NLP applications, including word sense disambiguation, word segmentation, semantic role labeling, speech recognition and etc. In fact, co-location alone could yield a feasible NLP model provided that we can collect a large enough knowledge base of these relations with good sentence coverage and appropriate weighting schemes. The demand for such a knowledge base creates a challenging issue: To what extent can one automate such an acquisition and validation procedure? Several such strategies will be proposed in this talk.

　　Wen-Lian Hsu received a B.S. from the Department of Mathematics, National Taiwan University in 1973, a Ph.D. degree in operations research from Cornell University in 1980. He then joined Northwestern University as an assistant professor and was promoted to tenured associate professor in 1986. He joined the Institute of Information Science in Academia Sinica as a research fellow in 1989. Dr. Hsu's earlier contribution is on algorithms for perfect graphs and special intersection graphs such as planar graphs and interval graphs. Most of his publications appear in JACM and SIAM J. Computing. In the meantime, he has applied similar techniques to tackle computational problems in biology and natural language.

　　Dr. Hsu is the inventor of the popular Chinese Input Method GOING in 1993, which is now used by over 1 million users daily in Taiwan. Dr. Hsu’s lab has won many international contests in natural language systems, including: the 1st place in the NTCIR6 2006 CLQA Chinese Question Answering Contest; the 1st place in NTCIR6 2006 CLQA English-to-Chinese Cross-Lingual Question Answering Contest; the 1st place in NTCIR5 2005 CLQA Chinese Question Answering Contest; the 1st place in SIGHAN 2006 Word Segmentation Contest; the 2nd place in SIGHAN 2006 Named Entity Recognition Contest. He is particularly interested in applying natural language processing techniques to the understanding of DNA sequences as well as protein sequences, structures and functions and also to biological literature mining. Dr. Hsu received many awards from the National Science Council, including the Distinguished Research Award in 1991, 1994, 1996 and the Appointed Distinguished Research fellow Award in 2005. He received the first K. T. Li Research Breakthrough Award in 1999, the IEEE Fellow in 2006, and the Teco Award in 2008. He has been the president of the Artificial Intelligence Society in Taiwan from 2001 to 2002. He is currently the Director of the TIGP Bioinformatics Program in Academia Sinica.