Data-model for voice search of agricultural information system

We present our early work in the area of voice-based information retrieval in an agricultural information system. The specific scenario deals with an information system that has the audio content in a local language, Kannada. End-users, mostly being low-literate farmers make phone calls to browse information about their crops. Since the crops are large in number, and audio being a sequential modality, this browsing process is often inefficient. We therefore propose to build a search system through which users could speak keyword which we foresee to be crop names* pesticide names* insect names* village names and such The challenge lies in retrieving the right audio information that matches the user query. The specific use case is helped by the fact that the content is available in its textual format. In this paper, we present a data-model that can be used for structuring the textual content to develop the index for efficient and improved retrieval through audio queries. We present the details of the scenario and explain the data and its nuances and then propose a data model to effectively utilize the structure for efficient information retrieval in this specific context. We believe that the proposed approach can have implications for other languages and other agriculture based information systems.