Data-to-text Generation with Variational Sequential Planning

Abstract We consider the task of data-to-text generation, which aims to create textual output from non-linguistic input. focus on generating long-form text, that is, documents with multiple paragraphs, and propose a neural model enhanced planning component responsible for organizing high-level information in coherent meaningful way. infer latent plans sequentially structured variational model, while interleaving steps generation. Text is generated by conditioning previous decisions previously text. Experiments two benchmarks (RotoWire MLB) show our outperforms strong baselines sample-efficient face limited training data (e.g., few hundred instances).