Implementation of skip-gram in PyTorch with following features:
- Negative Sampling
- Frequent Words Subsampling
- Batch Training
- Similarity Evaluation
- Analogy Evaluation
- Visualization
Provide corpus path and other parameters in config.py and run python train.py for training.
To evaluate generated word embeddings run python similarity_eval.py for similarity evaluation and python analogy_eval.py for analogy evaluation.
To visualize the embeddings run python plot.py.