MReD: A Meta-Review Dataset for Structure-Controllable Text Generation

TL;DR

The paper introduces MReD, a new dataset of meta-reviews with annotated categories to enable structure-controlled text generation, addressing the limitations of existing datasets in domain-specific controllability.

Contribution

It provides a large, annotated dataset for structure-controlled text generation and demonstrates methods to leverage this data for improved summarization and domain understanding.

Findings

Models can generate more structured summaries using the dataset.

Annotated data reveals insights into meta-review domain characteristics.

Challenges remain in fully controlling generation with the proposed methods.

Abstract

When directly using existing text generation datasets for controllable generation, we are facing the problem of not having the domain knowledge and thus the aspects that could be controlled are limited. A typical example is when using CNN/Daily Mail dataset for controllable text summarization, there is no guided information on the emphasis of summary sentences. A more useful text generator should leverage both the input text and the control signal to guide the generation, which can only be built with a deep understanding of the domain knowledge. Motivated by this vision, our paper introduces a new text generation dataset, named MReD. Our new dataset consists of 7,089 meta-reviews and all its 45k meta-review sentences are manually annotated with one of the 9 carefully defined categories, including abstract, strength, decision, etc. We present experimental results on start-of-the-art summarization models, and propose methods for structure-controlled generation with both extractive and abstractive models using our annotated data. By exploring various settings and analyzing the model behavior with respect to the control signal, we demonstrate the challenges of our proposed task and the values of our dataset MReD. Meanwhile, MReD also allows us to have a better understanding of the meta-review domain.