theamrzaki/text_summurization_abstractive_methods

Implements three distinct seq2seq architectures—bidirectional LSTM with attention, pointer-generator networks for hybrid abstractive-extractive summarization, and reinforcement learning-based decoding—across multiple languages (English, Hindi, Amharic, Arabic). Each model is packaged as self-contained Jupyter notebooks integrated with Google Drive and Colab's free GPU, eliminating local setup requirements. Evaluation uses a custom "zaksum" metric combining BLEU and ROUGE scores.

530 stars. No commits in the last 6 months.