Host: The Japanese Society for Artificial Intelligence
Name : The 27th Annual Conference of the Japanese Society for Artificial Intelligence, 2013
Number : 27
Location : [in Japanese]
Date : June 04, 2013 - June 07, 2013
In this paper, we first give a brief overview of three well known tasks in natural language processing: term extraction, named entity recognition, and keyphrase extraction. Then, we formalize our technical term identification problem as a matching from term forms to term types, that corresponds to surface representation of terms and technical concepts referred to by them. In order to deal with both polysemous and homonymous nature of terms, we use Japanese-English term pairs as basic semantic elements, and develop a statistical method to automatically annotate technical terms in text using both distributional and pattern-based features. Also, we show a practical implementation of a term identification server that specifically targets for scientific papers and discuss the issues needs to be addressed in future study.
