Spoken Web: Internet for the poorly literate

Faculty: Aaditeshwar Seth

Collaborators: Spoken Web team at IBM IRL

Students: Vishal and Jaiprakash

Project: How do we extend information services, such as those available through the Internet, to poorly literate people and those who have not had significant experience of interacting with computerized systems? Researchers at the IBM India Research Lab came up with the concept of designing the Spoken Web, in which people can dial into a voicesite through their phones to interact with voice applications. These voice applications may be like a discussion forum where people can leave questions and others can dial to answer these questions, or they may be transactional applications through which farmers can place orders for purchase of seeds or fertilizers, etc. We are working on a number of projects to enhance the Spoken Web vision.

• Develop an open-source one-box equivalent of the Spoken Web: Our goal is to make it easy for any community organization to be able to set up their own voicesite by just purchasing a computer, a telephony card, and a phoneline. We are using the Asterisk PBX server and the Voiceglue VXML browser to do this.
• Develop a centralized data-center solution of the Spoken Web: Once PSTN lines terminate in gateways, calls can be relayed over SIP to any Spoken Web instance deployed on the Internet. It is therefore feasible to run Spoken Web in a virtualized cloud-computing setup where new instances can be spawned off elastically depending upon the incoming call volume. This however leads to interesting problems on whether audio quality can be preserved in virtualized environments, how can scheduling algorithms be tailored to specific IO needs of voice based applications, monitoring the resource requirements for different voicesites, and how can control systems be put in place to pack different classes of voicesites on the same virtual or physical machine.