ADDRESS FEATURE WORD EXTRACTION APPARATUS, METHOD AND PROGRAM

摘要

PROBLEM TO BE SOLVED: To provide an address feature word extraction apparatus capable of widely extracting feature words not limited to facility names and capable of estimating the number of locations which appear in short documents using the feature words in a larger number than prior art.SOLUTION: An address feature word extraction apparatus of this invention: calculates a recognition degree of feature words in documents in document storage means which correspond to input locational information, the recognition degree in a specific area of a user group who are the creators of the documents, and associates the recognition degree with the creators and stores it in deviation storage means; calculates a ranking rise rate of the feature words covering the documents obtained from the document storage means on the basis of changes in an appearance probability of the respective feature words which appear in the documents in a narrow area within a wide area; calculates a score for the feature words on the basis of the recognition degree in the specific area obtained from the deviation storage means and the ranking rise rate of the feature words; and extracts words that have the top N highest feature scores and represent features of the input address, and attaches the input address to the extracted words and stores the extracted words in unknown address document storage means.