Text Summarizer

How to Use the Text Summarizer

1. Paste the text. Several paragraphs work best — the scoring compares sentences with one another, so a single paragraph gives it little to go on.
2. Choose a scoring method. TextRank picks the sentences most connected to the rest of the text; word frequency picks the ones densest in the text's commonest content words.
3. Set the length. Keep a fixed number of sentences or a percentage of them. Leave the first-sentence boost on for academic prose, where paragraphs usually open with their topic sentence.
4. Choose Summarize. The summary appears in reading order, the chosen sentences light up in the original, and a table shows how the top sentences scored.

How Extractive Summarization Works

The summarizer never writes a sentence of its own. It splits the text into sentences — knowing that the period in “et al.”, “e.g.”, “p. 45”, or “3.14” does not end one — gives each sentence a score, keeps the best, and prints them in the order the author wrote them. Headings are not candidates, and neither are sentences with fewer than three content words, which carry too little to score fairly.

TextRank

TextRank (Mihalcea and Tarau, 2004) treats each sentence as a node in a graph. Two sentences are linked when they share content words, and the link is weighted by the overlap: shared words ÷ (log n₁ + log n₂), where n is the number of distinct content words in each sentence. PageRank then runs over the graph with a damping factor of 0.85. A sentence scores well when it is linked to many sentences that also score well — when it is central to the text.

Word frequency

The frequency method follows H. P. Luhn's 1958 automatic abstracts. It counts every content word in the text — function words such as the and of are ignored, and plurals are folded into the singular — then scores each sentence by the average frequency of its words, relative to the most frequent word. Sentences dense with the text's main vocabulary rise to the top.

With the first-sentence boost switched on, the opening sentence of each paragraph has its score multiplied by 1.2 under either method. Scores in the table are relative: the best sentence is 100.

Worked Example

Choose Try a sample to load four short paragraphs on memory consolidation, one of which contains a deliberately off-topic sentence about the weather. With TextRank, three sentences, and the boost on, the summary keeps the definition that opens the text, the sentence on the brain replaying the day's activity during deep sleep, and the sentence explaining that replay moves memories into the cortex. Switch to word frequency and the third choice becomes the practical advice in the last paragraph. Neither method picks the weather. The top keywords are sleep, used five times, and memories, used four.

What a Summarizer Can and Cannot Do

• It selects; it does not understand. Scores come from shared and frequent words, not meaning. A sentence central in vocabulary but minor in argument can be chosen, and a key point made once in unusual words can be missed.
• Extracted sentences can dangle. A chosen sentence that begins “This replay…” or “However,” leans on the sentence before it. Read the summary on its own and restore the context it needs.
• It suits expository prose. Articles, reports, and textbook chapters work well. Fiction, dialogue, bullet lists, and very short texts give it little to rank.
• The summary is still the author's wording. Use it to see what a text emphasizes, then write your own summary — or quote and cite any sentence you keep.

Frequently Asked Questions

Is this an AI summarizer?

No. It is a fixed algorithm: the same text with the same settings always gives the same summary, and you can see the score behind every choice. It runs entirely in your browser — nothing is sent to a language model or any other server.

What is the difference between extractive and abstractive summarization?

An extractive summarizer, like this one, selects sentences from the source and copies them unchanged. An abstractive summarizer writes new sentences that paraphrase the source — shorter and smoother, but able to state things the source never said. Extraction is less fluent but cannot misquote.

Should I choose TextRank or word frequency?

TextRank favors sentences that connect to many others, which suits articles that circle back to a central claim. Word frequency favors sentences packed with the most-used vocabulary, which suits tightly focused texts. Try both: a sentence that both methods choose is almost certainly important.

How long should a summary be?

For reading notes, 10% to 25% of the sentences usually captures the main line of argument; for a long chapter, a fixed handful of sentences is easier to scan. If you are writing an abstract, check the target journal's or style guide's limit — APA, for example, limits an abstract to 250 words.

Can I paste the summary into my essay?

Not as your own writing. Every sentence is the source author's, so using one means quoting it, with quotation marks and a citation. The better use is as a map: see what the text emphasizes, then summarize it in your own words and cite the source.

Why does the summary seem to jump between ideas?

Each chosen sentence is lifted out of its paragraph, so connecting sentences and pronoun references are often left behind. Raise the length, try the other method, or read the highlighted sentences in the original to see what links them.

Is my text uploaded or stored?

No. Splitting, scoring, and summarizing all happen in your browser, and the text is gone when you close the page.