Extractive Summarization
A principal approach to automatic text summarization in which key sentences or passages are selected directly from the source document to form the summary. This ensures full traceability to the source and avoids hallucination, making it widely used in high-fidelity domains such as law and medicine.
Timeline (last 90 days)
Extractive summarization directly selects verbatim content from the original text, ensuring that every concluding sentence can be traced back to the source document.
Extractive summarization is constrained to sentences from the source text and cannot perform linguistic reorganization and compression the way abstractive summarization can.
Graph-theory-based extractive methods require no large-scale training, offer strong interpretability, incur low computational cost, and do not suffer from hallucination problems.
An arXiv paper proposes reframing summarization as extractive sentence selection to address the problem of factual hallucination in generative summarization within biomedical and clinical domains.
In extractive summarization, every output sentence is copied directly from the input, eliminating generation-induced factual errors at the architectural level.
Extractive summarization sacrifices some linguistic fluency and information compression rate in exchange for the verifiability that every sentence can be traced back to the source text.
All Facts (6)
Extractive summarization directly selects verbatim content from the original text, ensuring that every concluding sentence can be traced back to the source document.
50%UnverifiedExtractive summarization is constrained to sentences from the source text and cannot perform linguistic reorganization and compression the way abstractive summarization can.
50%UnverifiedGraph-theory-based extractive methods require no large-scale training, offer strong interpretability, incur low computational cost, and do not suffer from hallucination problems.
50%UnverifiedAn arXiv paper proposes reframing summarization as extractive sentence selection to address the problem of factual hallucination in generative summarization within biomedical and clinical domains.
50%UnverifiedIn extractive summarization, every output sentence is copied directly from the input, eliminating generation-induced factual errors at the architectural level.
50%UnverifiedExtractive summarization sacrifices some linguistic fluency and information compression rate in exchange for the verifiability that every sentence can be traced back to the source text.
50%