Home › Forums › OpenEars › Does generating language model usually take a long time?

Tagged: 

This topic has 2 voices, contains 2 replies, and was last updated by  aerialcombat 133 days ago.

Viewing 3 posts - 1 through 3 (of 3 total)
Author Posts
Author Posts
January 6, 2012 at 10:36 am #8369

aerialcombat

calling [languageModelGenerator generateLanguageModelFromArray:...] with language size of around 50 takes about 10 seconds on my iPhone 4. Would you say this is pretty normal?

I did test with different language sizes and the time seems to be directly proportional to the number of words in the language, but I never thought it would take this long, and I’m wondering if it’s perfectly normal. I haven’t explored into how a language model is generated, so I can’t say much more, but I’m just curious. Thanks. :)

PS, oh I AM using words that are most likely not in American dictionary, could this be the explanation maybe?
January 6, 2012 at 11:22 am #8371

Halle

Yes, there are two methods for finding the pronunciation — the main method is very fast and is used for words which are present in the big CMU phonetic dictionary, and the fallback method is used for words which aren’t present and takes about 10x longer. If you’re curious about the specifics of it, it’s pretty well documented in the LanguageModelGenerator source.
January 6, 2012 at 7:42 pm #8372

aerialcombat

Thanks Halle. :)
Viewing 3 posts - 1 through 3 (of 3 total)

You must be logged in to reply to this topic.