Deep text matching

Text matching using several deep models.

Implemented models

• Baseline model: Word vectors averaging + Fully connected feed-forward layers.
• Deep Bi-LSTMs based model.
• RNMT+ encoder based model.
• Transformer encoder based model.
• Multi head attention based model.

Get started

• The main entry point is at train.py.
• Test with very small dataset using test.py before training.
• Grid search hyper-parameters with grid_search.py.
• Use apply.py to load trained model.