Easy access to information is one of the key factors contributing to the economic and social growth of modern societies in the fields of agriculture, health, nutrition, education, commerce and governance. Learning how to deal with a specific crop disease, where the closest market with the best current price for wheat is, how to make home-made oral rehydration solution – all are examples where a timely and correct answer can be immensely beneficial, and in some cases may mean the difference between life and death. In the field of community health, international public health experts have stated that "providing access to reliable health information for health workers in developing countries is potentially the single most cost effective and achievable strategy for sustainable improvement in health care".

However, most of the available means of information access such as print media are useful only for the literate members of the society. Other modes like television and radio are non-interactive and computers, although being interactive are not suitable for the major portion of the society that is either not familiar with its interface and usage or does not have access to it at all. Hence information access gets restricted to literate and affluent circles. One solution to this problem can be a telephone based speech interface for human-computer interaction. These systems can provide cost-effective and natural forms of information access for large populations in the developing world – both those who are completely non-literate, as well as the semi-literate, those that have difficulty in reading fluently.

"Telephone-based Speech Interfaces for Access to Information by Non-literate Users” is a joint effort of CRULP in collaboration with the LTI department of Carnegie Mellon University and the Agha Khan University. The goal of this project is to investigate the use of speech interfaces in a field-deployed system by providing easy access to medical information to lady health workers in Pakistan. This will be achieved by developing a telephone based dialogue system consisting of an Urdu Speech Recognition system and a Text to Speech system that can interact with the health workers to answer their queries. However, a dialog system is much more than an ASR engine coupled with a TTS engine: a dialog system needs to be able to mimic human conversation abilities by providing an intuitive conversation flow, detecting and correcting recognition errors, and giving feedback to the caller throughout the call.

The Text-to-Speech system required for this project has already been developed by CRULP. In addition to the problems of ASR (Automatic Speech Recognition) that are still present for English, the prime impediment towards the completion of this project is the lack of research and local language resources for URDU. On an abstract level a speaker independent automatic continuous (and spontaneous) speech recognition system for local languages and its further adaptation to telephone based interface is required as a first step towards achieving this goal.