Summarization Abstractive

Abstractive summarization is the task of generating new text that captures the meaning of a source document, paraphrasing rather than copying key sentences. Unlike extractive summarization (which selects existing sentences), abstractive summarization uses transformer models (BART, T5, Pegasus) to produce human-readable summaries that may contain words or phrases not in the original. This is closer to how humans summarize: you read a paper and write a summary in your own words, not by cutting and pasting key sentences. The challenge is ensuring the generated summary is factually consistent with the source and doesn't "hallucinate" facts.