Sinhala Arithmetic Problems Solver

Abstract

"Sinhala arithmetic problems are commonly used by students in their primary school years. These problems often take a significant amount of time to solve due to their complexity. Additionally, students are often unsure about the accuracy of their answers. Implementing a system to solve these problems would greatly benefit society by enhancing the country's education system. To develop a system for solving Sinhala arithmetic problems, the author collected requirements and designed the system. After completing these steps, the author created a dataset consisting of 493 Sinhala arithmetic problems, each labeled with its corresponding arithmetic operator. Subsequently, a Long Short-Term Memory (LSTM) model was developed to identify the arithmetic operator in Sinhala arithmetic problems, and it was trained using the created dataset. In evaluating the core functionality of the system, the author achieved an overall accuracy of 87% in correctly identifying the arithmetic operator.

This research addresses an important gap in the literature on Sinhala arithmetic problem solving, especially since previous studies did not cover all of the fundamental arithmetic operations, including addition, subtraction, multiplication, and division. Additionally, the study uses deep learning techniques, which are novel in the field, to improve performance above earlier systems that only claimed 76 percent accuracy for limited procedures. The method created in this study offers step-by-step solutions to improve learning in addition to supporting the identification of arithmetic operators. The model provides a useful tool for teachers and students both by utilizing NLP approaches specific to the Sinhala language. In the end, this work contributes significantly to the educational resources available for Sinhala speakers and lays the foundation for further advancements in language specific QA systems."