Looking for speech to text

I’ve spent the day searching for a speech to text module that will work in Asterisk. (Note: I am looking for “speech to text” not “text to speech”). I’d like to email the text of voicemails.

I’ve just spent a session trying to build Sphinx3, pocketsphinx, Speech::Recognizer::SPX and am not having much success. It doesn’t seem to be well maintained for general use.

I am running Asterisk on linux. Anything other than sphinx out there, preferable open source? Thx.

It is not opensource, but Lumenvox works pretty well. I recently compared it to Sphinx2 and Lumenvox seemed to have better results. That said, I’m not sure you’ll be able to get a good recognition with a grammar file the size of a dictionary :smile: It works well with a small defined set of words. Have you taken a look at jott.com?

I am getting the impression that it takes a large db of transcribed voice files in order to have a general purpose speech recognition system capable of converting voicemail to text. There are now a number of services available that work by having you forward your phone (busy/no answer) to their server and they take the message, transcribe it using their engine, then email you the audio and transcription. I have been using one of these, PhoneTag.com for several years.

This one VoiceCloud voicecloud.com/api.php has an api and looks interesting.

I’d been using phonetag.com for my voicemail (before I had my own asterisk voip service).

I had my home and cell busy/no answer forwarded to my phonetag.com number ($10/mon). Phonetag.com records the voice mail and transcribes it to text, and they do a nice job of it. Then they email me the text with the audio file attached.

Now that I use voip/Asterisk I wanted to replace my voice to text function with something driven by Asterisk. Since I haven’t found an easy solution to that I found that I can have Asterisk email my voicemails to phonetag.com and they do the transcription and email it back to me.

It works and I’m content with it until I figure out how to implement voice to text myself.