Wednesday, October 14, 2009

L versus R strikes again in Japanese English

Unfortunately these keep showing up. One would think a spell-checker would pick up most of these but they are still there… The first one is from the supermarket. It used to be wrapped around an apple. The second one is from a restaurant’s menu by a lake near Mount Fuji. Spot the problem? I don’t know if it is overcompensation in this case but this is something that happens much more rarely – an L for an R… Of course this one would not be caught by a simple spell-checker but something a bit smarter which aims to find L <-> R substitutions could pick it up.

Ah the researcher in me is flaring up again :P




Betty said...

Responding to the researcher in you:P - I wonder why MS spellchecker doesn't use a language model to catch these mistakes in word choices. At least I don't believe MS does. Does it make the spellchecker too computationally expensive to run on the fly?

Ulaş said...

I would imagine the spell checkers check just that - the spelling. Hence most of these are valid words, spell checker is happy. There is a deeper, grammar checker thingie as well but that one matches things like singular/plural agreement. This is more of a special case I would imagine.

Betty said...

Right, that's what I mean. These spellings are correct but for the wrong word choice. A bigram or trigram language model would figure that out. I find it strange that a spell checker does not use a language model to determine the target word to compare the spelling against. May be it's too computationally expensive to run on the fly.