Abstract
This study proposes a pattern matching algorithm for math expressions, focusing on their two-dimensional locational structures. The "two-dimensional locational structure" in this paper represents the structure in which symbols such as fractions and exponents are not lined up horizontally. We regard that the math structure is expressed with a nesting structure of strings in our algorithm. In this study, math expressions described by MathML (Presentation Markup) are dealt with. Presentation Markup allows us to describe math expressions with the same appearance in multiple ways. This causes pragmatic insufficiency of the retrieving function if it only treats the descriptive format recommended by MathML specification. By normalizing MathML data, the proposed algorithm is realized with high robustness.
