| 摘要 |
<P>PROBLEM TO BE SOLVED: To propose a technology allowing high-accuracy specification of erroneous notation inside an electronic document (text) and corresponding right notation even in a field (e.g. a medical field) wherein a morpheme-analyzing dictionary is not completed and sufficient morpheme analysis accuracy is impossible to obtain. <P>SOLUTION: Based on processing results of a morpheme analysis means 2 and a word likelihood calculation means 3, an erroneous notation correction candidate specification means by preparer 4 extracts a character string having low appearance probability only inside a report written by a specific doctor as a candidate of the right notation. An erroneous notation correction means by preparer 6 extracts a character string having high appearance probability only inside a report written by a specific doctor as a candidate of the erroneous notation to the respective obtained character strings. Similarity is calculated to all combinations of the respective character strings by a word similarity calculation means 5, and the erroneous notation correction means 6 specifies a combination having high similarity as an erroneous notation word and its correction word. <P>COPYRIGHT: (C)2009,JPO&INPIT |
