摘要 |
<P>PROBLEM TO BE SOLVED: To easily generate information extraction rules from a training corpus. <P>SOLUTION: A text input part 10 successively inputs sentences. A tree structure extracting part 11 applies syntax analysis to the inputted sentences and extracts a syntax tree. A tree structure display part 12 indicates a tree structure on a display part on the basis of information on the extracted syntax tree. An annotation imparting part 12 interactively imparts an annotation while using the indication of the tree structure. A regular tree structure expression extracting part 14 extracts a tree structure expression for indicating corresponding rules from the tree structure and the annotation and stores it in an information extraction rule storage part 15. The information extraction rules are used for extracting information and matched with sentences. <P>COPYRIGHT: (C)2005,JPO&NCIPI |