Interactive Natural Language Query Construction for Report Generation

Question answering is an age old AI challenge. How we approach this challenge is determined by decisions regarding the linguistic and domain knowledge our system will need, the technical and business acumen of our users, the interface used to input questions, and the form in which we should present answers to a user’s questions. Our approach to question answering involves the interactive construction of natural language queries. We describe and evaluate a question answering system that provides a point-and-click, webbased interface in conjunction with a semantic grammar to support user-controlled natural language question generation. A preliminary evaluation is performed using a selection of 12 questions based on the Adventure Works sample database.