The project aimed to develop machine learning techniques for character recognition of Tirhuta Lipi, a low-resource language primarily spoken in Bihar and Nepal. A diverse dataset of Tirhuta vowels was collected and preprocessed for training machine learning models. Different algorithms such as Sklearn and Tensorflow Keras were tested for character recognition, and their performance was evaluated. Overall, the project focused on developing techniques to improve language processing tasks for Tirhuta Lipi.
This project contributes to the growing field of natural language processing for low-resource languages and highlights the potential of machine learning in language-related tasks.
| Model | Accuracy |
|---|---|
| KNN (raw image) | 0.47 |
| MobileNet embedding + Decision Tree | 0.55 |
| SVC (raw image) | 0.61 |
| MobileNet embedding + Gradient Boosting | 0.86 |
| MobileNet embedding + KNN | 0.94 |
| MobileNet embedding + SVC | 0.95 |
| MobileNet embedding + Logistic Regression | 0.97 |
This project is prepared in partial fulfilment of the requirement for for the the bachelor’s degree in Electronics and Communication Engineering. First and foremost, We would also like to extend our sincere thanks to our passout seniors, friends, and family for their support and guidance throughout our research project. Their valuable inputs and feedback have been instrumental in shaping our ideas and improving our work.
We would also like to express our gratitude to the Department of Electronics and Computer Engineering, Pulchowk Campus, Tribhuvan University for providing us with the necessary resources and infrastructure to carry out this research.
Last but not least, we would like to thank our instructor, Basant Joshi, for his guidance, encouragement, and support throughout this project. His expertise and knowledge have been invaluable in helping us to achieve our research objectives.
Finally, we would like to acknowledge the contribution of all the authors of this research report, namely Sumit Yadav, Raju Kumar Yadav, and Prashant Bhandari, for their hard work, dedication, and collaborative efforts in producing this research.
Any kind of suggestion or criticism will be highly appreciated and acknowledged.