isabella232 / query-focused-sum

Official code repository for "Exploring Neural Models for Query-Focused Summarization".

Exploring Neural Models for Query-Focused Summarization

This is the official code repository for Exploring Neural Models for Query-Focused Summarization by Jesse Vig*, Alexander R. Fabbri*, Wojciech Kryściński*, Chien-Sheng Wu, and Wenhao Liu (*equal contribution).

We present code and instructions for reproducing the paper experiments and running the models against your own datasets.

Table of contents

• Introduction
• Two-stage models
• Segment Encoder
• Citation
• License

Introduction

Query-focused summarization (QFS) aims to produce summaries that answer particular questions of interest, enabling greater user control and personalization. In our paper we conduct a systematic exploration of neural approaches to QFS, considering two general classes of methods: two-stage extractive-abstractive solutions and end-to-end models. Within those categories, we investigate existing methods and present two model extensions that achieve state-of-the-art performance on the QMSum dataset by a margin of up to 3.38 ROUGE-1, 3.72 ROUGE-2, and 3.28 ROUGE-L.

Two-stage models

Two-step approaches consist of an extractor model, which extracts parts of the source document relevant to the input query, and an abstractor model, which synthesizes the extracted segments into a final summary.

See extractors directory for instructions and code for training and evaluating two-stage models.

Segment Encoder

The Segment Encoder is an end-to-end model that uses sparse local attention to achieve SOTA ROUGE scores on the QMSum dataset.

To replicate the QMSum experiments, or train and evaluate Segment Encoder on your own dataset, see the multiencoder directory.

Citation

When referencing this repository, please cite this paper:

@misc{vig-etal-2021-exploring,
      title={Exploring Neural Models for Query-Focused Summarization}, 
      author={Jesse Vig and Alexander R. Fabbri and Wojciech Kry{\'s}ci{\'n}ski and Chien-Sheng Wu and Wenhao Liu},
      year={2021},
      eprint={2112.07637},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2112.07637}
}

License

This repository is released under the BSD-3 License.