Text editing guides

Extract an Article Summary and Check Its Sources

·

An extractive summary selects sentences that already exist in the article. Use the word counts to estimate how much text was removed and compare the selected sentences with their neighbors when a date, reference or condition matters.

Article Summarizer

Compare the three extracts from the channel report

Select Load example, leave Short and Bullet points selected, then generate the summary. The sample has six distinct sentences and 50 words. The three selected sentences use 24 words, a 52% reduction. They describe the first check, the debris and the reopening in that order.

The selector favors sentences with words recurring across different sentences after excluding a short list of frequent function words. Set the article language before you generate to use its Unicode word and sentence boundaries.

Unicode Standard Annex #29, text segmentation

Bring the safety meeting into the shortlist

Enter safety meeting, inland in Focus phrases and generate again. The meeting sentence replaces the one about debris. The other two excerpts stay in place. Inland appears in the unmatched list because it never occurs in the sample.

Open Check source excerpts and find the meeting at offsets 135–185. These offsets count UTF-16 units from zero. The start points at Volunteers and the end sits just after the full stop.

Bring the safety meeting into the shortlist
SettingMiddle summary sentence
Short, no focus phraseA rising tide pushed debris into the channel.
Short, safety meetingVolunteers scheduled a safety meeting at the café.

Check the saved sentences before sharing

Change Summary format to Paragraph and generate a fresh result. The same sentences appear with a single space between them. Download TXT writes UTF-8, including the accent in café.

Read the lead-in and follow-up around any excerpt that starts with a pronoun or gives only one step of a process. Include the original article when the reader needs the omitted details.

Common questions

Why is the summary shorter than my selected limit?

The limit is a maximum. Duplicate sentences differing only in case or whitespace appear once. Punctuation-only fragments are skipped.

Does focus matching understand related words?

A focus phrase matches the same consecutive word tokens, ignoring case and equivalent Unicode accent encodings. Variants such as tides and tide remain separate.

Sources

Put it into practice.

Related guides