Sign in
Microsoft.com
 
Microsoft
 
 
 

locked
Spoken translation - pronounciation context RRS feed

 > 

  • General discussion

  • Discussion
    Sign in to vote
    0
    Sign in to vote

    Guys,

    How are you proposing to handle the pronouncation context of a word? I'll give an example. If you pass the english word "bow" to the speech API (and ask it to be provided in English) you get it back pronounced as in a bow from the waist. You may have meant bow as in a bow and arrow. Of course given the one word there's not much to go on. If I pass "bow and arrow" though, I would expect it to come back pronounced correctly, which it doesn't at present. This doesn't even consider translation to another language issues!

    Will you somehow be using context to decide how the pronounciation is done? Or will this be handled by a purely collaberative "human based" approach by someone correcting the phrase?

     

    Cheers,

    Greg

    • Changed type Greg Neilson Thursday, March 25, 2010 5:49 PM It's a discussion really
    Thursday, March 25, 2010 5:49 PM

All replies

  • Discussion
    Sign in to vote
    0
    Sign in to vote

    Hello Greg,

    in general, context is evaluated in speech synthesis. However, the "bow" example is not well covered with context for speech at this time. Translation is phrase-based, so the context will have a significant impact. The translation depends on having seen the phrase in the real world. Say "bow and arrow" has a higher chance to produce the hunting term for "bow" than the waist bend.

    Examples:
    Hunting with bow and arrow.
    He made a deep bow.

    Chris Wendt
    Microsoft Translator

    Tuesday, June 1, 2010 3:49 PM