Framework for sense disambiguation of mathematical expressions

Mathematical expressions are indispensable for describing mathematical concepts or models. Although these expressions are formal representations, they contain ambiguity, i.e., a single expression could be interpreted as having multiple meanings. This feature prevents the flexible use of mathematical expressions in computation. In this paper, we focus on symbol-level ambiguity and consider the problem of labeling semantic information to each symbol as a classification problem. Then we propose a framework for solving the problem with supervised learning.