INDEX GENERATION FOR RETRIEVAL

摘要

PROBLEM TO BE SOLVED: To evade the overlapped description of the same word in the positional information of the partial document to be expressed by the number of character (number of byte) and to generate indexes for which data quantity is remarkably reduced as compared with a conventional device. SOLUTION: The document inputted by a paragraph take-out means 3 is divided into plural partial documents, unique identification data is imparted to these partial documents and the partial documents are stored in a first table 4 by making the identification data of the partial documents and the positional information on the partial documents in the document correspond. The word extracted from the partial document by a word extraction means 6 is stored in a second table 7 by making the word correspond to the identification data of partial document from which the word is extracted. An index generation means 8 relates the storage information of the first table 14 and the second table 7 by the identification data of the partial document and the index in which the word is defined as a key is generated.