Conchylicultor/DeepQA

Implements a sequence-to-sequence RNN with encoder-decoder architecture using two stacked LSTM layers for training on conversational datasets. Supports multiple corpus sources including Cornell Movie Dialogs, OpenSubtitles, Ubuntu Dialogue Corpus, and custom formats, with optional pre-trained word embeddings for faster convergence. Includes both CLI and Django-based web interface with Redis for deployment, plus TensorBoard integration for monitoring training progress.

2,924 stars. No commits in the last 6 months.