Step-by-Step: Separating Planning from Realization in Neural Data-to-Text Generation

04/06/2019
by   Amit Moryossef, et al.
12

Data-to-text generation can be conceptually divided into two parts: ordering and structuring the information (planning), and generating fluent language describing the information (realization). Modern neural generation systems conflate these two steps into a single end-to-end differentiable system. We propose to split the generation process into a symbolic text-planning stage that is faithful to the input, followed by a neural generation stage that focuses only on realization. For training a plan-to-text generator, we present a method for matching reference texts to their corresponding text plans. For inference time, we describe a method for selecting high-quality text plans for new inputs. We implement and evaluate our approach on the WebNLG benchmark. Our results demonstrate that decoupling text planning from neural realization indeed improves the system's reliability and adequacy while maintaining fluent output. We observe improvements both in BLEU scores and in manual evaluations. Another benefit of our approach is the ability to output diverse realizations of the same input, paving the way to explicit control over the generated text structure.

READ FULL TEXT

page 11

page 14

page 17

research
09/22/2019

Improving Quality and Efficiency in Plan-based Neural Data-to-Text Generation

We follow the step-by-step approach to neural data-to-text generation we...
research
02/04/2021

Data-to-text Generation with Macro Planning

Recent approaches to data-to-text generation have adopted the very succe...
research
09/26/2020

Learning to Plan and Realize Separately for Open-Ended Dialogue Systems

Achieving true human-like ability to conduct a conversation remains an e...
research
05/24/2019

Designing a Symbolic Intermediate Representation for Neural Surface Realization

Generated output from neural NLG systems often contain errors such as ha...
research
06/28/2023

You Can Generate It Again: Data-to-text Generation with Verification and Correction Prompting

Despite significant advancements in existing models, generating text des...
research
06/10/2021

AGGGEN: Ordering and Aggregating while Generating

We present AGGGEN (pronounced 'again'), a data-to-text model which re-in...
research
12/08/2021

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 langu...

Please sign up or login with your details

Forgot password? Click here to reset