wb14123/seq2seq-couplet

Built on TensorFlow, this seq2seq implementation learns to generate contextually appropriate second lines (lower couplets) given an input first line, using an encoder-decoder architecture with attention mechanisms trained on paired classical Chinese poetry. The project provides both a training pipeline with BLEU score evaluation via TensorBoard and a deployable web service, with Docker support for containerized inference and examples demonstrating generation of semantically coherent couplets.

5,506 stars. No commits in the last 6 months.