Mathematical expression recognition device and formula recognition method

摘要

PROBLEM TO BE SOLVED: To realize an OCR system capable of recognizing a numerical expression with high accuracy from a document including the numerical expression.;SOLUTION: In a numerical expression detecting part 113, an optimum path for connecting words is searched while selecting one of a text and the numerical expression for every work on the basis of the formal grammar and an evaluation value of the text and the numerical expression calculated on each word and a numerical expression area is detected. In a following numerical expression recognizing part 114, the horizontal positional relationship, the subscript character relationship, the superscript character relationship are determined by using a plurality of scatter diagrams different in the kind of before and after characters. Not only the partial relationship between the characters is determined, but alto the optical path can be decided from link candidates produced every inter-character, and the inter-character suffix relationship is determined by searching the optimum path under the consideration of a global evaluating condition.;COPYRIGHT: (C)2002,JPO