Although I wasn’t able to attend SpeechTek last week in New York, I did talk to a number of people who were there, along some of the vendors who were exhibiting. Based on the show floor and announcements this month, the industry has made a lot of headway into automated customer support, mobile applications, and entrance into emerging markets, such as unified communications. Here is fraction of what I missed.
In the automated customer service/support arena (in particular, coupled with AI), Voxify introduced their Call Steering Manager, which I blogged about on 8/21/07. SpeechCycle expanded their offerings with Level One Foundation – automated customer/technical support solutions for wireless and satellite companies, Internet-enabled consumer electronics, gaming devices, and hardware and software manufacturers. This expands their original virtual agents in the technical support arena that I wrote about in my April 2006 Innovations column in Speech Technology Magazine. SpeechCycle’s initial customer base focused on triple play technical support for cable TV, broadband Internet and VoIP digital, offering fairly in-depth and complex “virtual agent” technical support. I’m really happy to see that success in that area has allowed them to expand into new ones. Finally, Resolvity, who I didn’t get a chance to talk to yet, exhibited their automated customer support applications, and spoke in a session on “Artificial Intelligence in Voice Self-Service Applications”.
Syntellect debuted Voiyager (See 8/9/07 blog). The debut evidently went over really, really well, with moments of jaw dropping silence, or “I get it” moments as people related to the problems it solves. I still haven’t seen it, but I have high hopes of talking to a Voiyager customer soon to hear about their experience.
IBM WebSphere Voice Server Feature Pack for Speaker Verification (blissfully shortened to Speaker Identity Verification or SIV) was announced and is way cool. Speaker verification isn’t new, and still comprises a small subset of speech applications out there (I would like to see this rectified), but the IBM WebSphere Speech Solutions group has added some long needed functionality in this niche that will help to drive adoption, rather than pilots.
IBM added four primary enhancements. SIV is now text independent, which means that the users can enroll (create a voiceprint) by saying anything rather than reading from a script, and can be verified by saying anything as well. This really helps up the confidence level of verification because the user can be double verified by asking them any number of verification questions, such as account number, etc., coupled with the voiceprint, and it doesn’t limit the application with asking for one specific thing. IBM also introduced a policy manager as a services offering that complements SIV. The policy manager adds a dynamic question and answer dialog to the caller interaction, which increases security by combining a voiceprint with specific information asked of the caller.
Second, SIV is now language independent, which means that users can enroll in the system in any language and be verified in another. This means that a person using an application can also switch between languages with no impact to the verification. You might think that this is no big deal, but this is a big deal when you see the final two features they added, which are Speaker Change Detection and Speaker Tracking. SIV now monitors the entire call for assurance that the verified speaker answers all the questions, and SIV detects when the speaker has changed. So, the application can now tell if someone has called in and verified themselves, and has handed off the telephone to someone else. This coupled with no language limit means that the verified speaker can switch between languages if they desire without the system thinking another person has gotten on the call.
IBM sells SIV strictly through OEM partners such as Avaya, Nortel and Genesys, along with ISVs, so I hope to hear more deployment stories from some of them. One valuable thing that IBM has done for their partners is to create a Lab Services group that helps partners with proof of concepts, pilots, and other developments. A chargeable service, partners can either use the Lab Services group to train their own speech people, or just to get help with customer deployments.
I wish I could have been there to see more. By the way, I heard they had red phone booths that attendees could go into and use a phone to create a voice print, and then go into another one to test it out.