-
Notifications
You must be signed in to change notification settings - Fork 15
Neural Machine Translation for Sumerian and English
Ravneet Punia edited this page Nov 30, 2019
·
9 revisions
The project aims to build and train a neural network-based encode-decoder architecture for Sumerian-English Machine Translation in order to support experts in cuneiform studies with automated translations.
Ravneet Punia - Student Developer
Niko Schenk - Mentor
- Exploring dataset more closely, for better preprocessing, removal of duplicate phrases, dividing into the test, train, and validation.
- Implementation of the neural network-based encoder-decoder framework for Sumerian - English machine translation.
- Experimenting with different word embeddings, Word2Vec, GLoVe embeddings.
- Also exterminated with pre-trained embeddings from Wikipedia corpus. Archived better performance then learned from the dataset itself.
- Implemented transformers for Neural Machine Translation task, with exact configuration as suggested in the paper by Google.
- Implemented custom 2 layer encoder-decoder model for Neural Machine Translation task. Archived better performance then transform Model.
- Calculated BLEU score for every model architecture with tuning hyperparameter to boost the overall accuracy.
- Visualizing the attention activity of both models discussed above.
- Adding the command-line interface for language conversion
- Making a bidirectional model, Sumerian to English as well as English to Sumerian
- Publishing a research paper about the current implementation.