Marshall Harrison - "the gotspeech guy"

I'm not sure I like the sound of OCS absorbing Speech Server. OCS requires Active Directory integration and therefore really limits what I can do as an ISV. We deploy Speech Server applications as an appliance to various enterprises. If we are required to integrate into their ADS then it just won't work for us.

And am I to understand that I'm going to loose VXML and remain stuck in VS 2005?

I'm glad one of us is excited.

My suggestion to them would be to post the VXML source to CodePlex and let the community help out. This would a) ensure that those of us interested in VXML support continue to have it available and b) make it much easier to port VXML from other platforms as we could customize the stack to add support for features not found in the strict VXML 2.1 specifications. We use a lot of property tags in our Nuance implementations that are not available to us in the Speech Server world for example.

Unfortunately if ADS is a requirement I'm going to be hurting. Having just spent 2 months trying to get read-only access to one customer's ADS I can just imagine what I'd be looking at with OCS...

Any word on whether R2 will have support for dictation ?

Does this mean that the new announced support for direct SIP trunking in OCS R2 is not supported in Speech Server?

Thanks to the watchful eye of Elan - he spotted a press release about OCS 2007 R2. It sounds pretty cool - just check out the details: Dial-in audioconferencing - Office Communications Server 2007 R2 enables businesses to eliminate costly audioconf

Thanks for your feedback. OCS 2007R2 is an important step in Microsoft’s strategy to create a managed platform for SIP applications, yet for speech functionality clearly should be seen as an interim state. When Microsoft announced Speech Server’s integration into Unified Communications in August 2006, UC was still a concept that was being developed. Two years later it is clear that Unified Communications is the cornerstone to innovation in the telecom industry. I dare to say that IVR as a stand-alone business is on the decline. Microsoft strongly believes in the power of speech, and heavily invests in it, yet we do also believe strongly it should be regarded as a feature. A key feature, yet part of a bigger whole.

For Speech Server (2007) Microsoft postponed the tighter integration into OCS on demand of our installed base. And in OCS 2007 R2 Microsoft has clearly decided that there is more to be gained for our customers of speech being a part of the UC platform than there is to be lost not having a stand-alone IVR. That is where our investments will be.

In R2 the new speech platform is a work in progress. The accompanying speech tools are missing, Microsoft is still working hard on our application server platform, OCS can do better on integrating OA&M and reporting of third party SIP applications. That is why Speech Server (2007) is still supported for the R2 timeframe.

Creating GRUUs in Active Directory indeed is less than smooth. Yet the UCMA is an endpoint API. And AD offers a lot of infrastructure (like for security) which we choose to leverage. The key here (and partners, please jump in!) will be to build better tools than the sample code for a tool we provide in R2.

On VoiceXML. Microsoft recognizes that VoiceXML support is another key feature going forward. Our core strategy for the UC platform is to offer an Unified Communications Managed API and Web Services APIs. VoiceXML is a higher level API on top of that UCMA. The UCMA is capable to host an VXML browser on top of it. We think Microsoft will need to do some basic work to integrate VXML on top of UCMA and as a Workflow Activity in the UC Workflow Activities. Yet I will explore if we can put that code on CodePlex. Thanks for that suggestion.

In R2 the tighter integration into the Windows Workflow Foundation has been key. In the coming months Microsoft will be revealing more about ‘Oslo’ and Visual Studio’s future. That is a big bet of Microsoft where I think UC needs to be an integrated part of. Unfortunately it means that the roadmap is defined along those time lines and will take its sweet time... ‘Oslo’ needs to be ready before we can do our next push there.

We will be targeting a broad set of developers that are not necessarily speech developers. We want to take speech mainstream. As a feature of productivity enhancing applications, where speech can play its role. In this sense we are working on a dual strategy: the speech tools will be more low level. The application authoring tools will need to be more high level, and need to make UC application development a lot easier leveraging the Workflow Foundation work Microsoft is doing overall.

Some of you have been asking for more core access to the speech engines. MRCP 2.0 as offered by Aumtech is a great example of how currently speech can be used by other platforms. UCMA 2.0 does offer API level access to the speech engines, ASR and TTS. The speech investments in Microsoft are unprecedented and growing. We have more people working on speech than ever before. We are working on support for 26 telephony languages (12 being ready by R2). We are working on integrating the latest research insights into our product to make core speech technology even more robust. Including server side dictation, be it that is not ready yet (did you check out the Dictation Resource Kit download on Microsoft Download for Windows Vista?). Microsoft has the ambition to offer the best engines out there. Yet as a feature of a bigger developer platform story. Not as a stand-alone business.

On the telephony integration side there is a lot happening. Support of the legacy PBX/switch industry for SIP over TCP is growing. Also SIP Trunking specs are being finalized in the SIP Forum and Microsoft hopes some key carriers will offer SIP Trunking based on that standard. Speech Server (2007)’s tested gateways have been of a Dialogic and Audiocodes make, and Aculab’s boards. We depend on vendors to qualify against our product. Like Paraxip/Sangoma did. The same will be true for SIP Trunking. It will be terra incognita. Likely gateways will be your best bet for the year to come. That is just reality. Clearly with tens of millions of OCS licenses sold versus the amount of Speech Server deployments, OCS will be the focus for carriers to test against and Speech Server (2007) will be not their priority. Another reason why a integration of speech into OCS will be a good thing, leveraging the scale and growth of Microsoft’s Unified Communications as a whole.

I hope all the above gives a better insight in our current thinking.

Keep the feedback coming!

Albert Kooiman - Microsoft

This posting is provided "AS IS" with no warranties, and confers no rights.

I'm unclear - will I be able to work on my existing Speech Server apps in VS 2008, or only on new UCMA apps in 2008?  Thanks.

Great post with very insightful comments. I'm curious whether the transposing extends to the transposition of entire audio conferences as well. This way we wouldn't need to have someone write down details but could later on just look at who said what.

Please keep the discussion going. OCS R2 is THE application. I'm totally excited about it's features and need all the insights I can get to convince our management to go for it.

I'm currently working out an approach to using Nortel's CS1000E with along with OCS R2. If someone has any comments, I'd highly appreciate it. Thanks.

Since the current speech tools did not get propagated up to VS2008, how long do you think Microsoft will support them?  Looking at Micheal Dunn’s post

http://blogs.msdn.com/midunn/archive/2008/10/15/speech-server-2007-vs-ucma-v2-0-wf-activites.aspx

UCMA 2.0 does not seem to be a solid solution for IVR’s.

Any thoughts or suggestions?

Well, I'am now a little bit confused about the new R2 release. Does it mean that they substitute the "Speech Server application development technique" (define grammars with the editors, debugging applications with the SIP debugger, using the tuning and analysis tools, ...) with the UCMA 2.0 API (I've also read the article from Michael Dunn about that topic)?

Isn't this a step backwards? Allthough, I'm new in that business, I selected the Speech Server and its development tools and API, because it is a modern kind of VUI development (state-of-the-art?).

So what they are doing now? Skipping VXML support is also a bad decision, because it is a kind of de-facto metadata standard for voice applications (however, I prefered managed code applications). Okay, supporting SIP trunks is very cool (that costs me a lot of time before, to configure a IP-PBX working with Speech Server). Furthermore, VS 2008 integration is good and SALT skipping is aceptable (why we need another standard?).

Nevertheless, right now, for me are more negative items on my list as positives.

Cheers zazi

Thanks a lot for that significant answer Marshall. So it is just a better decision to wait for switching to the next version. Let's celebrate now the R2 launch ;)

Cheers zazi