Text Summarizer (Extractive)
Pick the key sentences of any article — your own words, never rewritten.
Or open a .docx, .pdf, .odt, .rtf, .html, .md or .txt file. Nothing leaves your device.
Summary
Every sentence above is taken word for word from your text, in its original order.
Keywords
Your text, with the chosen sentences marked
Every sentence by TextRank score
| Rank | Score | Sentence |
|---|
About the Text Summarizer (Extractive)
Paste an article, report, essay or chapter — or open a Word, PDF, OpenDocument, RTF, HTML or Markdown file — and get its most important sentences, chosen with TextRank (Mihalcea & Tarau, 2004), the graph-ranking method behind many summarisers. Pick the length as a number of sentences or a percentage of the text; the summary keeps the sentences in their original order and lists the text’s top keywords.
This is an extractive summariser: it only selects sentences you wrote, word for word, so it never invents facts or changes meaning. Each chosen sentence is marked in your text with its rank, and a table shows every sentence’s score so you can see why it was picked. It works for English and 17 other languages, offline, on your device.
How to use it
- Paste your text, or use Open file (.docx, .pdf, .odt, .rtf, .html, .md, .txt). Nothing is uploaded.
- Choose the length: a number of Sentences, or % of text (the share of the text’s sentences to keep).
- Check Language of the text — the right stopword list matters. The page suggests one when your text is in another script.
- Read the summary, or switch Show the summary as to bullet points. The marked text below shows where each sentence came from.
- Copy summary, copy the keywords, or download both as a .txt file.
Examples
An article on rooftop solar: how the system works, net metering, what affects its output, sizing it from your bills and warranties.
A typical system has panels on the roof, an inverter that turns their direct current into the alternating current used at home, and a meter that records what flows to and from the grid. The amount of power a system produces depends on the size of the panels, the direction and angle of the roof, shade from trees or nearby buildings, and the weather. They show how much power the household really uses, which decides the size of the system and therefore its price. Keywords: system, panels, household, power, roof, electricity, home, warranties, year, metering
3 of 13 sentences, 84 of 258 words. The third sentence begins with “They” (the electricity bills of the sentence before it) — a reminder that an extractive summary can lose the context of a sentence.
Common uses
- Getting the gist of a long report, article or chapter before deciding to read it in full.
- Students picking out the key sentences of their reading material to make notes from.
- Writing an abstract, a meeting brief or a newsletter blurb from sentences that already exist.
- Finding the keywords of a page for tagging, indexing or SEO.
How TextRank picks sentences
Every sentence becomes a point in a graph. Two sentences are linked by their word overlap, Similarity(Sᵢ, Sⱼ) = (number of words they share) ÷ (log |Sᵢ| + log |Sⱼ|), where |S| is the number of words in a sentence — dividing by the logs stops long sentences from winning just by being long. A weighted PageRank then scores each sentence by how strongly it is linked to other well-linked sentences:
WS(Vᵢ) = (1 − d) + d × Σⱼ [ wⱼᵢ ÷ Σₖ wⱼₖ ] × WS(Vⱼ), with d = 0.85,
repeated until no score changes by more than 0.0001; d = 0.85 is the damping factor the paper uses, after Brin and Page. The highest-scoring sentences form the summary, put back in their original order. Shared words are counted after removing stopwords (the, of, and …, from the stopwords-iso lists); English words are also reduced to their Porter stem (M. F. Porter, 1980), so “market”, “markets” and “marketing” match. Lines of fewer than four words, such as headings, are never chosen.
How the keywords are found
TextRank ranks words the same way: each content word is a point, linked to the words next to it, and PageRank scores them. The top third (between 5 and 20) are kept, and keywords that stand next to each other in the text at least twice become a key phrase (“solar panels”). The list shows the ten best with how often each occurs.
Choosing the length
For a news article or blog post, 3–5 sentences usually carries the main points. For long reports, try 10–20%. The readout shows how many words the summary keeps and how much shorter it is than the original.
Limitations
- Extractive only: sentences are chosen, not rewritten or shortened, so the summary can read unevenly and a chosen sentence may refer back to one that was left out (“They…”, “This…”).
- The method looks at shared words, not meaning: it cannot tell a key claim from a repeated aside, and works best on texts with clear, complete sentences — less well on lists, tables or chat logs.
- Word stems are used for English only; other languages are compared by whole words. The paper’s part-of-speech filter for keywords is replaced by stopword lists, so an occasional verb can appear among the keywords.
- Up to 4,000 sentences (a long report or a few chapters) are summarised at a time; split longer texts into parts.
- Scanned PDFs have no text layer: make them searchable with OCR PDF first. Old .doc files must be saved as .docx.
Privacy
Everything happens in your browser. What you enter or open here is not uploaded or stored by MySmartCoPilot. Opening a PDF downloads the PDF engine and an English word list from MySmartCoPilot once; your text and files never leave your device.
Frequently asked questions
Does the summarizer rewrite my text?
No. It picks whole sentences from your text and shows them exactly as you wrote them, in their original order. It never adds words or facts, which also means it cannot paraphrase or shorten a sentence.
How is this different from an AI summary?
An AI model writes new sentences and can get facts wrong; this tool only selects your own sentences with a published, repeatable method (TextRank), so the same text always gives the same summary and nothing is invented.
Why does a summary sentence start with “They” or “This”?
Because it was chosen without the sentence before it, which it refers to. Choose a slightly longer summary, or read that sentence in context in the marked text below the summary.
Which languages does it support?
English, Hindi, Marathi, Bengali, Gujarati, Urdu, Spanish, French, German, Portuguese, Italian, Dutch, Indonesian, Russian, Turkish, Arabic, Chinese and Japanese, each with its own stopword list. Sentences and words are split with the browser’s Unicode rules.
Can I summarise a PDF or Word document?
Yes. Open file reads the text layer of PDFs and the text of .docx, .odt, .rtf, .html, .md and .txt files on your device. The text appears in the box, where you can trim it before summarising.
Is my text uploaded?
No. The summary is computed in your browser and the page works offline once loaded. Opening a PDF downloads the PDF engine once; your file is read on your device.