Text Similarity Checker
Everything runs in the browser; your text is never uploaded
Enter text and press process. Everything runs locally; nothing is uploaded
How to Use
📖 Tool Introduction
Work out how much two texts overlap, by character or by word, and list what appears in one but not the other.
Deciding how much two passages share sounds simple and depends entirely on what you count. Comparing character by character suits Chinese and short strings, and reflects how close the two are on the surface. Comparing by word, taking English words and individual Chinese characters as the units, says more about whether the content overlaps than about whether the letters do. Neither is more correct, they answer different questions, and knowing which one you are asking is most of the work.
The shared part is counted as a multiset intersection, so a term appearing twice in both texts counts twice, and one appearing twice on one side and once on the other counts once. Simply checking whether a term is present would flatten that, and the difference matters when a text repeats itself, which is exactly the case where a similarity figure is most likely to mislead.
The denominator is the larger of the two sizes, and that choice is what makes the number behave. Against an average or against the smaller side, a short passage entirely contained in a long one scores low, even though anyone would call it fully overlapping. Dividing by the larger side gives that case the high score it deserves, and it also means the figure reads as the share of the longer text that the overlap accounts for.
Alongside the percentage, the tool lists what is unique to each side. That list is usually more useful than the score, and it is worth saying why: the number tells you how much changed, while the listing tells you where, and the second question is the one you can act on. A similarity figure on its own invites an argument about the threshold, which is rarely the argument worth having.
Everything runs in the browser, so what you paste never leaves your device.
One caution about how the number gets used. A similarity percentage invites the question of what counts as too similar, and that threshold is not something a tool can supply, because it depends entirely on why the comparison is being made. Two drafts of the same document should score very high. Two independent answers to the same question should not, and the figure will say so without explaining it. Treat it as a measurement and read the listing before drawing a conclusion.
✨ Key Features
- Compare by character or by word
- Multiset intersection, so frequency differences are not flattened
- Denominator is the larger side, so containment scores as it should
- Lists what is unique to each text
- Shows both lengths and the similarity
- Runs locally: nothing is uploaded
🚀 How to Use
- Paste the two texts to compare on the left
- Pick characters or words on the right
- Press process, or use Ctrl plus Enter
- The result box gives the score and a summary, with the detail below
- Copy result puts the summary on your clipboard
⚠️ Things to Note
- Data source: Similarity is computed in the browser with no external service
- Not semantic similarity: This measures surface overlap, not whether two passages mean the same thing
- Rewriting lowers the score: A paraphrase shares fewer words and scores lower, which is inherent to surface comparison
- Chinese by word means by character: Chinese has no spaces to segment on, so in word mode each character is a unit
- Privacy: Everything runs in the browser. What you paste is never sent to a server