📚 Glossary

Text Summarization

Text summarization is the process of producing a condensed version of a longer text that retains its key information, main arguments, and essential meaning.

Definition

Text summarization is the task of automatically or manually producing a shorter version of a document that preserves the most important content. Summaries reduce length while maintaining the essential meaning, key points, and — for narrative content — the logical structure of the original.

Types of Text Summarization

Extractive summarization selects and combines the most important sentences directly from the source text. The output consists of original sentences — nothing is rewritten. Extractive systems are simpler to implement but can produce choppy, context-lacking summaries.

Abstractive summarization generates new sentences that express the source's meaning in fewer words. Modern LLM-based summarizers use this approach, producing fluent summaries that may phrase information differently than the original. This is closer to how humans summarize.

Summarization vs. Paraphrasing

Summarization Paraphrasing
Length Shorter than source Similar to source
Goal Condense key information Re-express in different words
Detail Selective Complete

Quality Criteria for Summaries

A good summary: - Covers the main argument and key supporting points - Omits minor details, repetition, and examples that aren't essential to the main point - Preserves the logical structure and relationships between ideas - Avoids introducing new information not present in the source - Reads as a standalone document (doesn't require the reader to have read the original)

AI Summarization Quality

LLM-based summarizers produce high-quality summaries for structured, factual content. Common failure modes include: - Hallucination — introducing facts not present in the source - Flattening nuance — removing important qualifications that change the meaning - Missing the main argument — summarising supporting points while omitting the thesis

For important documents, human review of AI-generated summaries is recommended.

Paraphrasing · Large Language Model · AI Summarizer