Unsupervised Extractive Summarization

Unsupervised extractive summarization aims to automatically create concise summaries from text data without relying on labeled training examples, focusing on selecting the most important sentences. Current research emphasizes improving sentence representation learning through techniques like graph embeddings, contrastive learning, and incorporating document-level context, often employing graph-based ranking algorithms or Siamese networks. These advancements are significant because they enable summarization in low-resource settings and for diverse document types (e.g., medical notes, legal cases, social media), improving information access and efficiency across various applications.