Does Structure Matter? Leveraging Data-to-Text Generation for Answering Complex Information Needs

In this work, our aim is to provide a structured answer in natural language\nto a complex information need. Particularly, we envision using generative\nmodels from the perspective of data-to-text generation. We propose the use of a\ncontent selection and planning pipeline which aims at structuring the answer by\ngenerating intermediate plans. The experimental evaluation is performed using\nthe TREC Complex Answer Retrieval (CAR) dataset. We evaluate both the generated\nanswer and its corresponding structure and show the effectiveness of\nplanning-based models in comparison to a text-to-text model.\n