Open-domain clarification question generation without question examples
This addresses the challenge of enabling machines to communicate more effectively with humans by resolving ambiguities in natural language, though it is incremental as it builds on existing captioning methods.
The paper tackled the problem of generating clarification questions in dialogue to resolve misunderstandings, proposing a model that uses expected information gain to produce polar questions from an image captioner without supervised data, and demonstrated improved communicative success in a 20 questions game with synthetic and human answerers.
An overarching goal of natural language processing is to enable machines to communicate seamlessly with humans. However, natural language can be ambiguous or unclear. In cases of uncertainty, humans engage in an interactive process known as repair: asking questions and seeking clarification until their uncertainty is resolved. We propose a framework for building a visually grounded question-asking model capable of producing polar (yes-no) clarification questions to resolve misunderstandings in dialogue. Our model uses an expected information gain objective to derive informative questions from an off-the-shelf image captioner without requiring any supervised question-answer data. We demonstrate our model's ability to pose questions that improve communicative success in a goal-oriented 20 questions game with synthetic and human answerers.