WAV: Voice Access to Web Information for Masses

One of the main reasons for a large section of the world population to be left out of the internet revolution is, limited or no access to a computer due to economic, educational, cultural and age factors. Enabling masses to extract information from the web via voice will bring the Internet revolution to additional billions of people. In this paper, we describe a system called WAV (Web Access via Voice), that is a step in this direction. Departing from the traditional approaches of manually building a VoiceXML based site, the WAV system uses information from existing web sites to serve the user. Challenges to overcome include extracting contextually relevantinformation from the user and also from the pages returned by websites, reducing amount of information relayed to user over phone and maintaining the context of the conversation for easy renement based on feedback from the user. Our prototype system not only shows successful integration of many dierent technologies such as automatic speech recognition, scripts for web navigation, text to speech conversion, but also introduces a novel way of extracting information from web via voice in a programmatic manner. We describe initial solutions developed to tackle above challenges and demonstrate the feasibility of the system by describing proto type implementations on two popular web sites in India.