TellMe, which Microsoft bought two years ago, is rolling out an upgrade to its call center automation software which should improve its speech recognition rates. It is also adding Global Crossing as partner for reselling its VoiP carrier service, along with AT&T and Verizon. TellMe handles 2.5 billion calls a year for customers such as American Airlines and ETrade. Even a one percent improvement in automated call completion rates translates into millions of dollars a year for large call centers.
TellMe will be deploying a new text-to-speech engine with an almost-sexy female voice called Zira. She only sounds slightly robotic. Here is a clip you can download:zira_businesslisting
Another set of technologies can break up sentences into their constituent parts so that if the software doesn’t understand something it can ask for only the piece of missing information instead of repeating the entire question. For instance, if you say you want to fly from New york to San Francisco on Wednesday, and it got everything but the day, it would only ask you what day you want to fly instead of making you repeat your entire itinerary.
TellMe is adding a few other technology improvements under the hood as well, such as statistical models for predicting the next word you are going to say to narrow down the possibilities and acoustic modeling that adapts to your accent or speech pattern.
Finally, consumers will begin to see some of this speech-to-text technology in Windows Mobile 6.5. It will be able to read text messages in that Zira voice, and it will also make voice commands an input method for mobile phones running the WinMo operating system, which should be out some time in the fall.
First off, I am glad it’s getting better. But I have to say that until it gets REALLY GOOD (Ever tried to order airline tickets on a cellphone in a subway.. yeah, good luck) I will use a company with a local and real person over a call center-data service even if it costs me money!
I should probably add that I am 33 years old, and very tech savvy, I just can not stand having to yell your choices in some made up oxford English holding the phone 2 ft from your face.
Personally I appreciate personal service from knowledgeable people a lot, and am willing to pay for it.
Same with emails, chats and what ever contact surface it may be, and I don’t think I am alone.
Daniel
https://spideroak.com – Online Backup and File and Folder Sync
This is a good step for TellMe. Especially acoustic modeling” is one thing it had problems with earlier. In fact all such software lag behind in this area. But I do not know whether its next-word-recognition thing will work. We have to see how far it goes to guess the next word.
I’m looking forward to the speech to text technology and voice commands in WinMo. Btw, I heard Zira speak. Verdict: 55% robotic; 45% husky.
Better than the old Bethany voice.
Hey all,
Daniel at spideroak is just spamming TechCrunch with a push for spideroak. He has comments everywhere … crying for attention …
TellMe is the best acquisition Microsoft has made in the past 5 years
I agree, a solid, profitable acquisition. In my opinion, these are the types of services/companies Microsoft should be looking at acquiring. The allow MS to sell more services to existing clients. I think Microsoft should pattern themselves more like IBM and perhaps Oracle than Google. Their strengths lie with Enterprise customers.
I think I am encountering caching issues with Varnish on your server? I am trying to post from another browser. I will try from my mobile client if this one fails.
Asterisk is a free alternative to this.
asterisk.org
You can buy the RJ45 card for $50-100 on ebay
I say that because with askterisk you can buy a little intel atom dual core box and have your own system with something like vonage $25 a month service.
At any rate, I am leaving for the next week so wish me luck. My first stop is land of the Nikkei
I forgot to mention you can get an RJ11 to RJ45 splitter for about $5-10 if you google for one(RJ11 is the normal phone jack you need to plug into your PCI card). It’s not really worth having a SaaS system if you are larger than a couple people. PBX became practically free with varnish. I was using Kall8 when I was just a small business and their virtual PBX was only $2 a month including a toll free number. So there is a lot out there.
Sorry, PBX became practically free with Asterisk, not varnish. Varnish is an at times inferior HTTP proxy. The poor man’s load balancer.
Oh come on, if Zira sounds sexy to you, then you must be having your 9 second sexual thought at that moment. “Sexy”? Wow, it sounds like my mother.
No, it doesn’t sound sexy. It sounds “almost-sexy.” They are trying to make the voice sexy, but you can tell it’s a replicant, er, software.
To the folks talking about Asterisk… her name is Allison Smith and you’ve heard her before.
http://www.theivrvoice.com/
You just might not have realized it at the time.
http://www.thei...m/html/demo.htm
Now -that- is vocal talent.
Specifically… the IVR Demo http://www.thei...m/html/demo.htm
I wish all these automated voice systems would go bankrupt. It’s impossible to talk to a real person anymore before listening to 10 mins of automated voice systems, which almost never understand the speech correctly.
Agree with you Mik. But automated voice systems make things easier for people (at least for those who run the business) and solves delays in call waiting. Do we not have people to talk around that we look for someone real to answer our call for inquiry of something?
solves delays in call waiting
This is absolutely not true.
You should just go to http://www.gethuman.com/ which lists all the bypass mechanisms for various IVR and tree menus of the call center world for a variety of companies.
I want my voice to be used for Automated Calls. How can i apply to Microsoft for this.
You must be kidding when u said “She only sounds slightly robotic”
If that is true, this might be good. Voice recognition is really good , but this thing is all about SPEECH recognition?” Wow.
I think all these automated voice systems would go bankrupt. It’s impossible to talk to a real person anymore before listening to 10 mins of automated voice systems, which almost never understand the speech correctly.
If you want an alternative to TellMe i suggest trying voxeo (www.voxeo.com) or their new product Tropo (www.tropo.com) which makes making inbound and outbound calls very easy.
Also the TTS and ASR is pretty much best of breed and very good.
Paul
Speech enhances customer loyalty and emulates dialogue with callers. Good workflow design and VXML skills will help in recognizing the caller and predicting what they are looking for. For mobile and hands-free callers, it is a very convenient method of data exchange.
We need less call-center automation and more cool, sexy mashups in this area. Voice and speech are such phenomenal technologies that, with the exception of some recent mobile and dictation applications, suffer sorely from the role they play in the IVR/customer service sector.
Nortel had a directory assistance speech app in the 1990s, but it took GOOG411 for anyone to get excited about it. “[Breaking] up sentences into their constituent parts” has been around for a while too.
I hope someone can use technology by MSFT/TellMe, Voxeo, Nuance etc and do something exciting with it. Call center automation is where the money is, I suppose. But the breakthrough will be elsewhere.
Checkout http://www.vooices.us – Its inbound voice that can control javascript that anyone can code againgst using the API.
Tropo (http://www.tropo.com) is that breakthrough you are looking for. Develop directly in Ruby, Groovy, Python, Javascript or PHP. No need for VXML or CCXML in that case.
Almost sexy female voice think “not” sounds more like the voice you hear on COPS “Robbery at night shop 14/15 south Louis Avenue suspect is on foot armed and dangerous ” .
Ask me to repeat myself one more time Tell me.
I don’t think even a sexy female voice can conceal the fact you have to say the commands more then once on a occasion the frustration remains.