PURPOSE: To improve not only the retrieval efficiency regardless of slightly erroneous read of each character in a document but also the retrieval precision.
CONSTITUTION: A prescribed number of, for example up to three candidate characters are selected as the read character regardless of slightly erroneous read of each character in the document, and a registered keyword whose first character coincides with one of these candidate characters is searched, and there is a high probability that this coinciding character is correct. Characters following the document character corresponding to this first character are read in order and are subjected to the same processing as the first character to recognize characters in corresponding succeeding places of a keyword to be retrieved. Respectice candidate characters or changed characters in places other than the first place at the time of coincidence with corresponding constituting characters of the registered keyword 4 are denoted as new examples of erroneouly read characters, and erroneous read history data is updated based on them.
JP3127869 | SYSTEM AND METHOD FOR EXTRACTING SIMILAR DATA |
JP2002024455 | LAWYER INTRODUCING SYSTEM |
JP2005301584 | SERVER, METHOD AND PROGRAM FOR DISTRIBUTING SUMMARY ARTICLE |
KINOSHITA HITOSHI
FUJI FACOM CORP