Solving Explicit Arithmetic Word Problems via Using Vectorized Syntax-Semantics Model

Abstract

This paper presents an algorithm for solving explicit arithmetic word problems by using and fusing new and effective methods. The proposed algorithm is based two novel ideas. First, it bases on the newly built syntax-semantics method. This method is very effective because it significantly reduces the difficulty caused by the variety of semantics expressions. Second, it uses a vector computing method to enhance the syntax-semantics method. The proposed algorithm consists of four steps. It first encodes the problem text into a vector sequence. Second, it generates the matching candidates of the problem. Third, it computes matching scores between the vectorized models and the candidates. Finally, it decodes quantity relations and their positions in the original problem text. Experimental results show that the proposed algorithm outperforms the baseline algorithms on the benchmark datasets.