Text Summarizer

This text summarizer shortens articles, reports, essays and meeting notes by selecting the sentences that carry the most weight, rather than rewriting anything. It uses extractive summarisation: every sentence is scored on how often its words appear across the whole passage, with a small bonus for sentences near the opening, and the top-scoring sentences are returned in their original order. You paste the text, pick a summary length of 20, 30, 50 or 70 percent, and receive the summary along with the original word count, the summary word count, the compression percentage and how many sentences were kept. Because the output is made only of your own sentences, nothing is paraphrased or invented, which suits students skimming source material, researchers triaging papers and anyone deciding whether a long document is worth reading in full. The passage needs at least 50 characters and more than one sentence.

The summarizer first splits your text into sentences by looking for a full stop, question mark or exclamation mark followed by whitespace. It then builds a frequency table of every word in the passage after removing a list of common stop words such as the, and, of, is, to, with and would, so that filler words cannot inflate a score.

Each sentence receives a raw score equal to the sum of the frequencies of the words it contains. That raw score is then multiplied by a position weight calculated as 1 minus the sentence's position in the document divided by the total sentence count, times 0.3. In practice the first sentence keeps its full score and the final sentence keeps about 70 percent of it, which reflects the habit of stating the main point early.

The sentences are ranked by weighted score, the top portion is taken according to your chosen ratio, and those sentences are sorted back into their original order before being joined. Nothing is reworded, so the summary reads as a set of direct quotations from the source.

Suppose you paste a 300-word background section containing 12 sentences and leave the summary length on Short (30 percent).

The number of sentences to keep is calculated as the sentence count multiplied by the ratio, rounded up: 12 times 0.3 is 3.6, which rounds up to 4 sentences. The tool scores all 12, ranks them, and takes the four highest. Imagine those turn out to be sentences 1, 3, 7 and 10; they are placed back in that order so the summary still flows forwards through the document.

If those four sentences contain 96 words between them, the results panel reports: original length 300 words, summary length 96 words, sentences 4 of 12, and compression 68 percent, because 1 minus 96 divided by 300 equals 0.68.

Switching the same passage to Very Short (20 percent) gives 12 times 0.2 equals 2.4, rounded up to 3 sentences, and a correspondingly higher compression figure. Switching to Long (70 percent) gives 12 times 0.7 equals 8.4, rounded up to 9 sentences.

Treat the summary as a highlight reel, not a rewrite. Because sentences are lifted verbatim, any figure or claim in the summary is exactly as the author wrote it, which is useful when accuracy matters more than smooth prose. The trade-off is that pronouns can lose their referent: a kept sentence beginning "This meant the trial was halted" may no longer sit next to the sentence that explained what "this" was.

Frequency scoring rewards repetition, so texts that hammer one term repeatedly summarise well, while texts that make a single decisive point in one short sentence may see that sentence dropped. Dialogue, bullet lists without full stops, and passages with abbreviations such as e.g. or Dr. can also confuse the sentence splitter, since it breaks on punctuation followed by a space.

Practical habits help. Strip navigation text and reference lists before pasting, keep paragraphs intact, and start at 30 percent before moving up or down. If the compression figure looks extreme, check the sentence count rather than the word count, since the ratio applies to sentences.