Nicky: Toward a Virtual Assistant for Test and Measurement Instrument Recommendations
Robert H. Kincaid, Graham Pollock · 2017
Natural language question answering has been an area of active computer science research for decades. Recent advances have led to a new generation of virtual assistants or chatbots, frequently based on semantic modeling of some broadly general domain knowledge. However, answering questions about detailed, highly technical, domain-specific capabilities and attributes remains a difficult and complex problem. In this paper we discuss a prototype conversational virtual assistant designed for choosing test and measurement equipment based on the detailed measurement requirements of the test engineer. Our system allows for multi-stage queries which retain sufficient short-term context to support query refinement as well as compound questions. In addition to the software architecture, we explore an approach to ontology development that leverages inference from reasoners and minimizes the complexity of entering the specifications for a large collection of instruments. Finally, we provide insights into the issues of building this system and provide recommendations for future designs.