Solving Arithmetic English Word Problems Using Intermediate Representation

Developing a system to tackle math word problem solving is recognized as a complex undertaking that necessitates machine common sense, including logical inference. Despite the ongoing endeavors devoted to this field, developing a solver system that can effectively handle diverse datasets continues to pose a significant challenge in the present times. This project pioneers the evaluation of the Encoder-Decoder approach as a means of constructing an English language math question-answering system. Unlike methods requiring human feature engineering, this approach has displayed promising potential as a solver for the Chinese language. In our research, we customized the model for the English language, performed assessments using a subset of the MathQA English datasets, and carried out a sequence of experiments with the goal of improving the model's precision and scope.