Speech to text

Help
2007-12-19
2013-05-30
  • Nobody/Anonymous

    Hi There,

    I'd like to know if MARF is able to transcript speech to text.

    Thanks

     
    • Serguei A. Mokhov

      Hi,

      > I'd like to know if MARF is able to transcript speech to text.

      Not directly at this point, sorry. You can try CMU Sphinx.

      We are working on speech-to-text plug-ins, but not quite
      there with the actual man power contributing... :( I am
      considering making Sphinx's stuff to make as plugins for
      MARF for this purpose as the task by itself is pretty
      complex and time consuming to do right, especially if
      you want a general-purpose speech-to-text engine. Specific-purpose
      is easy, but limiting.

      -s

       
    • Nobody/Anonymous

      I am also looking for such a program. Really, what we need is something that will convert simple sounds to phonemes. Not interested in language processing.

      Ex. If someone said "papa" (onto an attached microphone) the program should generate the text 'papa'.

      We are working on the front end of an language independent phoeneme processing system.

      Please respond to

      siyengar@winona.edu

       
    • Serguei A. Mokhov

      Date: Thu, 7 Feb 2008 12:32:34 -0500 (EST)
      From: Serguei A. Mokhov <mokhov AT cs concordia ca>
      To: siyengar AT winona edu
      Cc: Serguei A. Mokhov <mokhov AT cs concordia ca>
      Subject: Re: [marf-dev] [marf - Help] RE: Speech to text

      Hi,

      On Wed, 6 Feb 2008, SourceForge.net wrote:

      > Date: Wed, 06 Feb 2008 11:10:54 -0800
      > From: SourceForge.net <noreply@sourceforge.net>
      > To: noreply@sourceforge.net
      > Subject: [marf-dev] [marf - Help] RE: Speech to text
      >
      >
      > I am also looking for such a program. Really, what we need is something
      > that will convert simple sounds to phonemes. Not interested in language
      > processing.
      >
      > Ex. If someone said "papa" (onto an attached microphone) the program
      > should generate the text 'papa'.

      This task (of 'papa') is not difficult and can be done without phoneme
      processing. Unless I misunderstood you, you only need Segmentation on
      word-boundary (by detecting silence gaps) and matching up the segmented
      word utterances to a training set dictionary (preprocessed and with
      features extracted). Such word-boundary segmentation for MARF is being
      worked on as I am writing this.

      > We are working on the front end of an language independent phoeneme
      > processing system.

      If you have to to phonemes, which are components of each word's
      utterances, the simple word-boundary segmentation is not enough, and
      especially in a continuous flow of speech data followed by a state-machine
      based on hidden Markov models (HMMs) and the like. Then you get a good
      semblence of a full speech-to-text syste. MARF due to lack of man power to
      work on that doesn't do that. (You are welcome to contribute it though if
      you are up to it and want to help out ;).

      As I mentioned before, CMU Sphinx (also hosted on SourceForge.net and is
      also in Java and also licensed in BSD) does that (HMMs and others), so you
      can look at that engine.

      As far as MARF concerned, as earlier discussion in this and similar
      threads indicate, we may simply adapt Sphinx' code as a set of plug-ins
      for MARF to do he speech-to-text tasks.

      Hope that is of any help to you.

      > Please respond to
      >
      > siyengar@winona.edu

      Doing that ;) With a copy to the forum...

      --
      Serguei A. Mokhov            |  /~\&nbsp;   The ASCII
      Computer Science Department  |  \ / Ribbon Campaign
      Concordia University         |   X    Against HTML
      Montreal, Quebec, Canada     |  / \&nbsp;     Email!

       
      • Nobody/Anonymous

        Dear Mokhov:

        I responded to your msg at mokhov@cs.concordia.ca - let me know if you received my response.

        Iyengar

         
    • D G

      D G - 2008-03-02

      Iyengar,

      I was wondering if you were able to find a solution to the generic speech-to-text transcription problem.  I'd appreciate some pointers etc.

      Thanks.
      -- Dmitry

       
      • Serguei A. Mokhov

        Dmitry,

        Dr. Iyengar and I exchanged a couple of private emails,
        and then he fell silent. Either very busy or found
        an alternative, e.g. the earlier mentioned CMU Sphinx
        (also hosted on SourceForge). Take a look at Sphinx
        it may do most of what you seek.

        -s

         

Log in to post a comment.

Get latest updates about Open Source Projects, Conferences and News.

Sign up for the SourceForge newsletter:





No, thanks