Improve Query Focused Abstractive Summarization byIncorporating Answer Relevance (QFS)

This is the implementation of the paper:

Improve Query Focused Abstractive Summarization byIncorporating Answer Relevance. Dan Su, Tiezheng Yu, Pascale Fung Findings of ACL 2021 [PDF]

If you use any source codes or datasets included in this toolkit in your work, please cite the following paper. The bibtex is listed below:

@misc{su2021improve,
      title={Improve Query Focused Abstractive Summarization by Incorporating Answer Relevance}, 
      author={Dan Su and Tiezheng Yu and Pascale Fung},
      year={2021},
      eprint={2105.12969},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

Abstract

Query focused summarization (QFS) modelsaim to generate summaries from source docu-ments that can answer the given query. Mostprevious work on QFS only considers thequery relevance criterion when producing thesummary. However, studying the effect of an-swer relevance in the summary generating pro-cess is also important. In this paper, we pro-pose QFS-BART, a model that incorporatesthe explicit answer relevance of the source doc-uments given the query via a question answer-ing model, to generate coherent and answer-related summaries. Furthermore, our modelcan take advantage of large pre-trained mod-els which improve the summarization perfor-mance significantly. Empirical results on theDebatepedia dataset show that the proposedmodel achieves the new state-of-the-art perfor-mance

Dependencies

Our experints enviroments is:

python 3.6, pytorch(v1.6.0), transformers(v3.0.2)

Also install other dependencies via

pip install -r requirement.txt

Experiments

Download Data

You can download the Debatepedia data via the original [link] (https://github.com/PrekshaNema25/DiverstiyBasedAttentionMechanism)

Released Checkpoints

We also released our pretrained model for reproduction.

• model

Training/Fine-tuning QFS model

sh scripts/finetune_debatepedia_qfs.sh

Evaluation

sh scripts/eval_debatepedia_qfs.sh