Abstract:
Sinhala is the mother tongue of most Sri Lankans. The number of Sinhala
speakers worldwide is around 20 million. Although many Sri Lankans speak
Sinhala fluently, many don‟t have a deep understanding of its grammatical
rules. Special knowledge of Sinhala grammar is required to solve and correct
grammatical errors. There are many unique features of the Sinhala language
that make it challenging to use it correctly. It is written in letters, and the
sentence structure is different from English (subject-object-verb order).
Many people find it difficult to apply the grammar rules they learn in the
field of education. This leads to common mistakes in written Sinhala.
Currently, there are no applications for checking Sinhala grammar in
handwritten Sinhala documents. Most existing grammar correction
applications require user-typed text and perform the check Sinhala grammar
itself only while the user is typing. It can be recognized that there is a
significant gap in the resources available for Sinhala to meet the unique
challenges posed by handwritten content in particular. Recognizing this
issue, this research aims to develop a mobile application that can effectively
detect grammar issues in handwritten documents. The proposed solution uses
optical character recognition (OCR) technology developed by convolutional
neural networks (CNN) to accurately recognize Sinhala characters in
handwritten documents. Also, the Natural Language Toolkit (NLTK) and
rule-based approach have been used to check the grammatical errors in the
identified Sinhala text. The model is trained using a dataset of 7, 0000
handwritten samples collected from various sources covering all Sinhala
characters. Apart from this, it is also possible to identify the mistakes in the
basic grammar of the Sinhala language by using the rule-based approach.
Using these technologies, this application enables all its users to improve
their writing skills and develop better language skills, even those who have
slight knowledge of grammar rules.
