🤖 Text Paraphraser

The tool first splits your text into sentences using a regular expression that looks for full stops, question marks and exclamation marks, keeping each terminator so punctuation survives. Each sentence is then tokenised into alternating word and non-word spans, which preserves your original spacing and punctuation exactly.

Every word token is tested against a series of filters before it can be replaced. Common function words on a stop list are left alone. Tokens of two characters or fewer are skipped. Words that are capitalised in the middle of a sentence are skipped as likely proper nouns, although fully capitalised acronyms are still eligible. Whatever survives is looked up in the synonym dictionary for the selected tone.

Strength controls a replacement probability applied to each remaining candidate: light gives each one a 22 per cent chance of being swapped, medium 48 per cent and strong 72 per cent. A synonym is then picked at random and matched to the original word's capitalisation.

Tone changes which dictionary is consulted. Formal and simple tones apply their own preference rules, so utilize becomes use under the simple tone. In strong mode the tool also varies the opening of roughly every third sentence.

Take this input: The company decided to improve the system to help workers finish important tasks quickly. That is fourteen words in total.

The filters remove four of them immediately. The, to, the and to are all on the stop list, so they are never candidates. That leaves ten content words: company, decided, improve, system, help, workers, finish, important, tasks and quickly. At medium strength each of those ten has a 48 per cent chance of being replaced, so on average roughly five of them change.

One possible run produces: The firm chose to enhance the system to assist workers complete important tasks rapidly. Six words changed: company to firm, decided to chose, improve to enhance, help to assist, finish to complete and quickly to rapidly. The rewritten percentage is calculated against the full word count, so six changes out of fourteen words displays as 43 per cent.

Run the same input again and you will get a different result, because both the probability test and the synonym choice are random. Switching to light strength would typically change two words out of the same ten candidates; strong would change about seven.

Because the engine works one word at a time, it has no model of meaning. Synonyms are not always interchangeable in context, so always read the result end to end rather than copying it straight out. Watch particularly for words with multiple senses: a swap that is correct in one reading can be wrong in yours, and technical vocabulary should usually be left intact even when a synonym exists.

Turn on the highlight option while you review. It marks every changed word, which reduces checking a long paragraph to scanning the highlights and confirming each one still reads correctly. Start at light strength and increase only if the output is too close to the source, since higher strengths compound the risk of an odd choice.

Be realistic about what this achieves. Substituting synonyms does not change the structure of an argument or the order of ideas, so it is a drafting aid, not a way to disguise copied work. Presenting someone else's writing as your own remains plagiarism however many words are swapped, and reworded text still needs a citation. For assessed work, check your institution's rules before using any rewriting tool.