Text Summarizer

Features

• Adjustable summary length from 10% to 60%
• Three scoring methods
• Keyword extraction showing what drove the selection
• Preserves original sentence order by default
• Runs entirely offline, no model, no API

How to use it

1. Paste the text you want condensed.
2. Set how much of the original to keep.
3. Read the summary and the key terms.
4. Copy or download the result.

Extractive summarisation, and why it is honest

This is extractive summarisation: it selects existing sentences rather than writing new ones. Each sentence is scored on the frequency of its significant words, after stop words are removed, plus a positional bonus, since the opening sentences of a document and of each paragraph disproportionately carry the main claims. The highest-scoring sentences are kept.

The advantage over the abstractive approach used by large language models is that nothing can be fabricated. Every sentence in the summary appeared verbatim in the source, so the summary cannot invent a statistic, misattribute a quotation or assert something the document never said. For legal, medical or technical material, that guarantee is often worth more than fluency.

The limitation is equally clear. Extracted sentences can read disjointedly, and pronouns lose their referents, a kept sentence beginning “This meant that…” may no longer have an antecedent. The method also works best on structured expository writing such as news, reports and documentation, and poorly on narrative, dialogue or anything where meaning is distributed rather than concentrated. Keeping original order, which is the default here, helps coherence considerably.