Saturday, December 07, 2013

[xogscuje] Typoable words

Find pairs of words with:
Short edit distance between them
Impossible to use context to determine which word is meant
Occur rarely enough to catch the writer unaware
Have significant consequences for swapping the two

This seems to be a hard AI NLP problem.  Hypothesize that language tends to evolve away from such pairs.  Are there examples of language evolving this way?

Closest I can think of is switching one name for another.

Update: larboard and starboard evolved to port and starboard.

2 comments :

Anonymous said...

One example:

It amazes me that in Polish the words for the numbers 9 and 10 are both 8 letters long and identical except for one letter in the middle:

9 = dziewięć (with "w" in the middle)
10 = dziesięć (with "s" in the middle)

Ken said...

Similarly, English has thirteen...nineteen and thirty...ninety.