| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide a keyword extraction device that extracts a keyword in a document in dependence on evaluation based on a corrected word frequency considering constituent words and on the number of constituent words. <P>SOLUTION: A digitized document is divided into words, which are classified as parts of speech and, if necessary, analyzed for basic word forms. According to the analysis results, keyword candidates are extracted by keyword candidate extraction rules describing patterns to be extracted as keyword candidates. The extracted keyword candidates are evaluated in terms of constituent words of the candidates, and according to the evaluation results, keywords are extracted from the keyword candidates. <P>COPYRIGHT: (C)2004,JPO |
