摘要 |
<P>PROBLEM TO BE SOLVED: To provide a method and device for detecting a boundary in a token, in which a boundary(sentence boundary) existing in a token column is precisely detected. <P>SOLUTION: This method for detecting a boundary in a token column includes using both chunking processing and successive dependency analysis processing, and executing boundary decision processing based on re-chuncking by adding new raw materials from those analytic results, and evaluating the boundary of a sub-cluster by chuncking processing, and directly applying results including the scores of the chuncking processing to successive dependency analysis, and executing the boundary decision processing by successive dependency analysis or evaluating the boundary of the sub-cluster by chuncking processing, and executing boundary decision processing by successive dependency analysis by applying only the sub-cluster of chuncking processing. <P>COPYRIGHT: (C)2008,JPO&INPIT |