Shotnest · Library
Notice a short word that changes a statement
Booking is active and Booking is not active overlap by 0.75, and only the word not separates them; read the one-sided tokens before trusting either text.
How one short word moves the result
The comparison splits both normalized texts into distinct tokens. Booking is active has three: booking, is, active. Booking is not active has four. Three are shared out of four in total, giving 0.75. The difference list shows not only in the second text, and nothing only in the first.
The opposite meaning costs only a quarter of the overlap. A longer sentence would hide the word even more: in a twenty-token notice, one added not would still leave an overlap above 0.9. The ratio measures shared words, not shared meaning.
Case and accents do not count as a change. BOOKING IS ACTIVE and Booking is active are equal after normalization, with an overlap of 1.
Try the original example
- First text
- Booking is active
- Second text
- Booking is not active
- Exactly equal
- No
- Equal after normalization
- No
- Token overlap
- 0.75
Compare two status messages
- Paste the earlier status first and the later one second.
- Read the Only in first and Only in second lists before the overlap figure.
- Look for negations and status words such as not, no, cancelled or expired.
- Confirm the current state with the provider that issued the message.
- Paste each message on its own; extra shared lines on both sides raise the overlap and make the difference easier to miss.
- Only in second
- not
- Word overlap in the app's duplicate check
- 0.75
- App duplicate word threshold
- 0.7
Where this matters in the app
Shotnest's duplicate check would treat these two texts as close enough on words, because 0.75 is above its 0.7 threshold. What keeps them apart is the image: a duplicate also needs the same width and height and a near-identical picture fingerprint, and a changed status line usually changes the picture.
- Assuming a high overlap means the second message only repeats the first.
- Reading the overlap in the wrong direction; the lists, not the figure, show which text has the extra word.
- Forgetting that the app drops one-letter words from its word sets, so a changed a or I never counts there, while the comparison tool keeps every token.
Similarity is not a verdict
The tool cannot tell which booking state is real. Keep both screenshots until you have checked, then tidy up.