The Tamil language is spoken by 80 million people around the world. The translation between Tamil and English leads to a significant impact by helping in the understanding of Tamil scripts, which otherwise would be a tedious, costly, and time-consuming process. Thus, developing an automated system to perform Tamil to English translation would save human time and effort. We publicly release a new high-quality corpus for standard training, evaluation, and report results experiments with two different architectures based on Encoder-Decoder to translate Tamil to English. We further tried to improve it by experimenting with pre- trained word embeddings and tuning hyperparameters. Although Google-Translator also provides Tamil to English and vice versa, our implemented architectures, along with the new dataset, completely outperformed the Google Translator with a margin of 7.5 BLEU score. Moreover, our proposed model solves out of vocabulary and polysemy problems up to a greater extent.
Feel free to download our paper using this link
Here are our trained models, as discussed in research paper: Link
In our experimentation we used 236,427 parallel English - Tamil Sentences, further we add more sentences and to the dataset. The final dataset is divided into six English <--> Tamil parallelly translated files for ease of distribution, please check the dataset folder. If you want to test our latest trained model and other languages we covered, please visit TranslateMe Network
If you still face an issue, feel free to contact the authors:
All the evaluated methodologies for Tamil-to-English translations were done using the same dataset for both the architectures. We compared our results with Google translator . The transformer translator outperforms the proposed Tamil translator and the Google translator, which is known to be the best-reported model for machine translation. While experimenting with Tamil translator, it became clear that pre-trained embeddings performing better than trained from Corpus, hence for transformer translator, only tested with pre-trained embeddings.
Apart from BLEU Score evaluation, human evaluation was done by 3 Tamil scholars, given 100 randomly chosen translation from each other, using the following rating scheme:
Final scores user rating is calculated as an average of all ratings given by scholars to each model, which is shown in below table.
| Model Architecture | BLEU | User Rating |
|---|---|---|
| 1. Google Translator | 8.6 | 2.6 |
| 2. Tamil Translator | 21.6 | 14.6 |
| 3. Tamil Translator + pretrained Embeddings | 14.9 | 3.1 |
| 4. Transformer Translator + pre-trained Embeddings | 16.1 | 3.8 |
If you use our dataset or any other part of the paper, don't forget to cite us using
@article{jain2020neural,
title={Neural machine translation for Tamil to English},
author={Jain, Minni and Punia, Ravneet and Hooda, Ishika},
journal={Journal of Statistics and Management Systems},
volume={23},
number={7},
pages={1251--1264},
year={2020},
publisher={Taylor & Francis}
}
