Ubona…voice search engine for Bangalore

by Amit Ranjan on November 8, 2007

Ubona - About Ubona_1194515707150.pngUbona is a voice search engine that has just gone live in Bangalore. All you need to do is to call their number (080-4070 0000) and speak out the name of any of the place or product (currently only for restaurants, hotels, sweet shops, bakeries). Their speech recognition system will comprehend what you have said, and connect you with your desired destination or, if you are calling from a mobile phone, send the information as SMS. The service is accessible from any landline or mobile phone at no extra cost. While they are starting off with food joints, they plan to soon extend this to include other domains- entertainment, health, travel, consumer electronics, gifts, shopping etc. Ubona is targetting not just the consumer market but also business users (as a way for companies to reach out to their consumers).

Voice search engines are not easy to build; there are issue of speech recognition, lack of audio clarity, accent differences etc, but if the system can be trained to work in the majority of cases, this can be a very compelling product. Specially for the mobile platform, where form factor considerations are a big bottleneck for search to pick up. BuzzWorks is another voice search engine that was featured in the last edition of Proto, though (if I remember correctly), they seemed more focused on local utilities (like plumbers, carpentars, car mechanics etc).

{ 10 comments… read them below or add one }

Gaurav November 9, 2007 at 4:35 am

This is really rocking… I tried this service and was really amazed with the recognition quality. Having tried my hands at Speech recognition a while back, i called this service with the apprehension that it will not work most of the times but to my complete surprise i got 7 correct recognition out of 8 times i tried. This is a really complex subject and it becomes even more challenging in the Indian Context with people speaking in various accents.. My Congratulations to these guys for having cracked this.. I’m sure if the technology works well, a lot of other useful services can also be launched… Keep up the good work guys…

amit November 9, 2007 at 11:06 am

Gaurav,

Being in delhi, I did not try out the service; glad you find it useful.

How does the system actually handle the incorrect recognition?

Gaurav November 9, 2007 at 8:01 pm

Hi Amit,
The one error that i got was when i said “Sukh Sagar” and it recognized as “Shubh Sagar”. …..the system asks u to press 1 to connect if the name was correct ,3 for getting an SMS and 4 if the name was recognized wrongly … so if the recognition is wrong u can press 4 and then speak the restaurant name again…
BTW, today i tried the system again just to play around cos i found the service really exciting…. and it was such a cool experience when i connected to Barbeque Nation in flat 20 secs… Man.. i hope u get this in delhi soon… BTW just for the kicks u can try it from Delhi also… and in case u don’t know too many restaurants in blr, try – Barbeque Nation, Aromas Of China, China Pearl, Blue Ginger.. and lemme know..

umesh November 10, 2007 at 6:27 am

Hi
This is Umesh from Ubona. Like to provide some clarification on the technology front. Speech recognition cannot really recognize what the user has said but gives the probability of what it can be from a dictionary by using statistical methods. The task becomes really complex when the size of the dictionary becomes larger.So when user says something the system gives some hypothesis of what the user could have said and expects the user to confirm that. The false trigger can happen if the word or the restaurant name is not in the dictionary, if it is said in a way other than way it is defined in the dictionary, if there is interleved speech because of other people talking at the background , extreme noise or in some cases because of the mismatch of the statistical models. We have more or less handled the last two conditions and we will be working to address the others with more and more people using the system.

spandana November 12, 2007 at 6:45 am

sounds very promising.

sms support is missing, as is comprehensiveness of data (even within food/restaurant segment), leading me to believe that most of the matching is pretty rigged, rather than a scalable algorithm or platform.

it would be interesting how this would monetize and find synergies/competition against just dial.

Umesh November 12, 2007 at 8:26 am

Dear Spandana,

Thanks for trying our service and giving your valuable feedback.

We do have SMS support, if the call is from a mobile phone in India (and not from a landline/overseas).

In terms of the comprehensiveness of data – we have covered almost all the restaurants, hotels, bakeries, cakeshops, pubs. It’s quite challenging to get information of businesses in India as there is no reliable/accurate source available as is the case in US. Our team is working towards making the data as complete as possible.

We have tested our platform with a large dictionary and have got remarkable results. I hope you appreciate the complexity of launching such a service in India with diverse dialects/accents. We are very confident of our technology and we will be launching services for other domains in very near future.

We really appreciate your feedback and would love to hear more of your thoughts.

PS April 17, 2008 at 10:31 am

Hi,
is this service still on? have you started something new on this?
Do you see voice as a potential business in mobile in the new future?

PS April 17, 2008 at 10:34 am

when do u launch pan india?

race April 23, 2008 at 9:59 am

Dera, Friend
Here alot of clssifides Like Accounting/Finance,Government,Electronics etc.
which is provided without charge to you, is a World Wide Web site on the Internet that is designed to allow

users to post classified ads and/or to view classifieds ads posted by other users or interact with those users.
Please Visit For More Detail
http://desidirectory.com/Bangalore/Listings.asp

Rahul Gupta September 16, 2008 at 11:03 am

Congrats Ubona Team for developing such a versatile system.
July have really lost big Hunks.

Leave a Comment

{ 1 trackback }

Previous post:

Next post: