Overview

This repository contains submission for the CS779 Machine Translation Competition (English → Hindi / Bengali). It reflects an engineering-focused approach to Neural Machine Translation, emphasizing iterative experimentation, empirical validation, and robustness to real-world noisy multilingual corpora — beginning with GRU/LSTM Seq2Seq baselines and culminating in a 4-layer Transformer encoder–decoder architecture enhanced with pretrained embeddings.

Competition Result Summary

Metric	Score
BLEU	0.175
chrF++	0.417
ROUGE	0.44

---

Repository Structure

word2vec-project/
├── data/
│   └── dataset.txt                 # Training corpus (if applicable)
├── notebooks/
│   └── dp-word2vec-t4_final.ipynb  # Main Word2Vec implementation notebook
├── requirements.txt                # Python dependencies
├── .gitignore                      # Git ignore file
└── README.md                       # This file