With improvements in speech recognition and natural language processing capabilities, voicebot systems show promise to run interactive information services for less-literate populations in developing regions. In this context, we describe our initial experiences towards building an automated question-answering system in the domain of sexual and reproductive health and rights. This system is trained on data acquired from an IVR (Interactive Voice Response) platform on which users could record questions, which were then moderated and sent to an expert to get answers. Our goal is to now use this data to build an automated answer retrieval system so that questions can be answered in real time by retrieving an appropriate answer from the corpus of questions and answers available so far. Our insights are likely to be useful for several initiatives using IVR systems and looking to automate their search and retrieval functionality.
Paper
Full text
Initial Lessons from Building an IVR-based Automated Question-Answering System
Semantic Scholar · Computer Science · 2020
Abstract
With improvements in speech recognition and natural language processing capabilities, voicebot systems show promise to run interactive information services for less-literate populations in developing regions. In this context, we describe our initial experiences towards building an automated question-answering system in the domain of sexual and reproductive health and rights. This system is trained on data acquired from an IVR (Interactive Voice Response) platform on which users could record questions, which were then moderated and sent to an expert to get answers. Our goal is to now use this data to build an automated answer retrieval system so that questions can be answered in real time by retrieving an appropriate answer from the corpus of questions and answers available so far. Our insights are likely to be useful for several initiatives using IVR systems and looking to automate their search and retrieval functionality.