When a tool tells you two names in your manuscript look similar, it's doing something specific and simple, and understanding it tells you what such a tool will and will not find. The technique is comparing strings to each other rather than to a dictionary, so it works on words that no dictionary contains.
Edit distance
The measure counts how many single-character operations turn one string into the other: insert a character, delete one, or change one. Two strings one operation apart are very close; four apart are probably unrelated. A tool picks a threshold and reports pairs below it.
| Pair | Distance | Likely |
|---|---|---|
| Aeryn and Aerynn | 1 | A variant |
| Kaelith and Kaelyth | 1 | A variant |
| Marden and Mardenn | 1 | A variant |
| Toren and Doren | 1 | Could be two characters |
| Ashcroft and Ashcraft | 1 | A variant, or two families |
| Vale and Vail | 2 | Ambiguous |
Rows four and five show the limit. Distance is a measure of string similarity, not of intent, and two genuinely different characters can be one character apart. The tool reports; you decide.
Why the threshold is a trade
- A tight threshold reports only near-identical strings and misses variants that differ by two characters.
- A loose threshold catches more variants and fills the report with unrelated short words.
- Length matters: one character difference in a four-letter word is proportionally large; in a twelve-letter word it is small.
- Case and punctuation can be included or ignored, which changes what is reported.
- Frequency helps. A pair where one appears ninety times and the other twice is almost certainly a variant.
Point five is the useful heuristic and it's why counts matter alongside similarity. Similarity says these two look alike; frequency says which one you meant.
Where this sits
The Novel Kit's duplicate and alias checker runs exactly this matching across every term in the manuscript and flags the near-identical pairs, with the frequencies that tell you which form is dominant. The decision about whether a pair is a variant or two characters stays yours, because it depends on the book.
The general reason a spellchecker cannot do this is in why a dictionary is not enough.
A practical next step
Run the checker over your manuscript and look at the flagged pairs. Where one form is frequent and the other rare, you've almost certainly found a typo.
Sources checked
- How to tell if two names are too similar
- Why spellcheck misses invented names
- Character names too similar
Frequently asked questions
What is edit distance?
The number of single-character changes needed to turn one string into another. Aeryn and Aerynn are one apart; Aeryn and Marcus are not close.
Why does the threshold matter?
Because it decides what is reported. Too tight and you miss variants; too loose and the report fills with unrelated words.
Is it the same as a spellchecker?
No. A spellchecker compares against a dictionary. This compares the words in your book against each other, which is why it works on invented names.