LSTM-Based Bi-Directional Sequence-To-Sequence Model for Solving Arithmetic English Word Problems

Solving arithmetic English word problems is a challenging task that requires the integration of natural language understanding and mathematical reasoning. In this research paper, we propose a solution using an LSTM-based bi-directional sequence-to-sequence model to tackle this problem. Our study focuses on the MathQA dataset, which provides a diverse set of arithmetic word problems. The LSTM-based model utilizes an encoder-decoder architecture, with the encoder processing the input word problem and the decoder generating the corresponding mathematical expression or answer. The bi-directional nature of the LSTM allows for the capture of contextual information from both past and future contexts, enhancing the model's understanding and performance. To evaluate our proposed approach, we conducted experiments on the MathQA dataset. We compared the performance of our LSTM-based model with other baseline models commonly used in similar tasks. Our results demonstrate that the LSTM-based bi-directional sequence-to-sequence model achieves superior performance in solving arithmetic English word problems on the MathQA dataset. The model showcases its ability to capture complex linguistic patterns and mathematical relationships, leading to accurate predictions. Furthermore, our experiments highlight the effectiveness of leveraging bi-directional information flow for enhanced contextual understanding and improved performance.