To improve a search speed and search accuracy in search of an electronic document including a table in the text.
In search for the electronic document based on a vector space method, when a search condition described as a farmland to be sold in Yosino-cho, Yosino-gun, Nara prefecture is inputted, for example, a document such as a selection chart, which includes words such as Yosino-gun, Yosino-cho, and a farmland but includes no information about the farmland existing in the Yosino-cho in Yosino-gun, Nara prefecture, may be erroneously determined as a suitable document. In this search method, each of cells inside a table in the document is assumed as one document, and plurality of document vectors equivalent in number to the cells are generated for the document shown in the figure. A distance between the document vectors and a search vector generated for the search condition is computed, and the document corresponding to the document vector is determined as a suitable document when at least one document vector gives a distance below a threshold value to the search vector. It is effective when a value inside the cell is comparatively long.
JP2001283220A | 2001-10-12 | |||
JP2000339347A | 2000-12-08 | |||
JP2002312370A | 2002-10-25 |
Next Patent: ELECTRIC METER MAINTENANCE MANAGEMENT WORK SYSTEM