Speaking of Standards

Just in time for the submission deadline for IETF 81, Voxeo’s Dan Burnett was able to get in what will hopefully be the last draft of the Media Resource Control Protocol (MRCP), version 2, a.k.a. “MRCPv2″. The new document (revision 25!) is available at:

http://tools.ietf.org/html/draft-ietf-speechsc-mrcpv2

If you aren’t familiar with MRCP, I explained a bit about the problem it is trying to solve back in March when the “Last Call” was issued for comments.

Somewhat predictably, there were comments after the “last call” and so with this revision 25 Dan B has incorporated many of those comments. You can see the diff of what has changed at:

http://tools.ietf.org/rfcdiff?difftype=–hwdiff&url2=draft-ietf-speechsc-mrcpv2-25.txt

Coming in at a whopping 226 pages, this is a significant document. Here’s hoping for Dan B’s sake that this is in fact the final draft!

This week the IETF issued the “Last Call” for comments on version 2 of the Media Resource Control Protocol (MRCP), a.k.a. “MRCPv2″. Voxeo’s Dan Burnett is the editor for the specification and this week released the 24th (and hopefully final) revision of the document:

http://tools.ietf.org/html/draft-ietf-speechsc-mrcpv2

I’ve written about MCRP before, in November 2008 and then in September 2009, but if you aren’t familiar with MRCP, it is a protocol that allows platforms such as our Prophecy platform to easily interact with Automatic Speech Recognition (ASR) or Text-To-Speech (TTS) engines. You can think of it like this:

MRCP is an open standard that enables your application platform to connect to any “MRCP-compliant” speech engine. This can allow you to connect to speech engines that, for instance, support a language that is not one that we commonly make available. Or you might have a very specialized speech engine for a specific use case. I recall one customer who was looking to do speech recognition in a very noisy environment and wanted to use a very specialized speech engine that they had access to. Because that speech engine was MRCP-compliant our platform could connect to it. That’s the kind of interop that MCRP can enable.

As the IETF Last Call announcement indicates, the IESG is asking for comments by April 13, 2011. After this Last Call process, the IESG will make a decision that will hopefully lead to MRCPv2 being published as an RFC. (If not, there could be another review process.)

As I’ve written about previously, the Media Resource Control Protocol (MRCP) is currently undergoing revision within the IETF to arrive at a new “MRCPv2″. Voxeo’s Dan Burnett has been editing the draft specification to incorporate the latest rounds of comments and last month released the 20th revision of the Internet-Draft:

http://tools.ietf.org/html/draft-ietf-speechsc-mrcpv2

If you aren’t familiar with MRCP, it’s a protocol that allows products such as our Prophecy platform to easily interoperate with Automatic Speech Recognition (ASR) or Text-To-Speech (TTS) engines. You can think of it like this:

With MRCP, your application platform can connect to any “MRCP-compliant” speech engine. It’s an open standard that we certainly like because it unlocks our platform and lets you use any of the great number of speech engines supported by Prophecy. We’ve also had customers approach us in the past about using special speech engines – and the open interface of MRCP provides the way in which this can happen.

In any event, MRCPv2 is moving closer to completion – if you have any comments about the latest draft, now is a really good time to send them in to the editors.

With IETF 73 coming up shortly in Minneapolis, those of us here in Voxeo were very busy last week getting our Internet-Drafts updated in time for Monday’s submission deadline. One of the major pieces of work was done by Dan Burnett with his new revision of the Media Resource Control Protocol Version 2 (MRCPv2) draft.

MRCP is actually a fascinating protocol to me (okay, admittedly, I’m a standards geek) in that it provides an open standard that allows a system to very easily interoperate with different “media processing resources” such as Automatic Speech Recognition (ASR) or Text-To-Speech (TTS) engines. This is how, for instance, our Prophecy product is able to easily use different ASR or TTS engines. In a very simplified view, it looks something like this:

where the “MRCP Client” is, in our case, Prophecy. Now the cool part about this is that if you need a specific ASR engine for a task, if you can find an “MRCP-compliant” engine it should be able to easily interoperate with Prophecy. Say, for instance, that you needed speech rec for a language we didn’t support, a special TTS engine or something like that.

Anyway, the new draft of MRCPv2 is out there and goes into this in an extraordinary amount of detail. If you do have any comments, by the way, Dan Burnett is open to hearing them (his email address is at the end of the draft).

P.S. If you’ve used our Realtime Debugger or our Prophecy Log Search feature inside of Evolution, you’ve no doubt seen a bunch of messages related to MRCP – this is all part of the communication between our main execution environment within Prophecy and the various ASR and TTS resources being used to execute your application.

Technorati Tags: MRCP, IETF, Standards, ASR, TTS, SIP

Hi, I’m Dan Burnett. I’ll be posting here occasionally about the speech-related standards in W3C and IETF.

I’m an editor of VoiceXML 2.0/2.1, SSML 1.0/1.1, and MRCPv2, an author of EMMA 1.0, PLS 1.0, SCXML 1.0, and the forthcoming VoiceXML 3, and a contributor to almost every other specification from the Voice Browser and Multi-modal Working Groups.