| 摘要 |
PROBLEM TO BE SOLVED: To appropriately acquire a topic word relating to an object. SOLUTION: An object extraction part 11 extracts an object from a document group; a topic word candidate extraction part 12 extracts topic word candidates from each document including the object; and a topic word candidate accumulation part 13 classifies each document into a document domain on the basis of the source of the document, accumulates document domain frequencies represented by the number of document domains in which a document including a topic word candidate is classified and document frequencies represented by the number of documents including a topic word candidate, and stores the document frequencies in a topic word candidate database 17. When an object for retrieving a document is input, a topic word candidate acquisition part 14 acquires topic word candidates corresponding to the input object with reference to the topic word candidate database 17, and a topic word candidate selection part 15 calculates a score S that becomes higher as the document domain frequencies are higher for each topic word candidate, and selects a topic word candidate whose score S is equal to or more than a threshold S<SB POS="POST">0</SB>. COPYRIGHT: (C)2013,JPO&INPIT |
