/SegMerge

One implementation of the paper "Exploiting Discourse-Level Segmentation for Extractive Summarization".

Primary LanguagePython

Introduction

One implementation of the paper Exploiting Discourse-Level Segmentation for Extractive Summarization, and the EDU level segmentation part is updated by using our new DMRST_Parser.

  • Please note that the SegMerge_Infer.py is only for building the EDU segments, not the summarization part.
  • Users can apply it to process the input text, and obtain the EDU segments, as well as the merged segments for extractive summarization.
  • This repo and the pre-trained model are only for research use. Please cite the papers if they are helpful.

Package Requirements

The model training and inference scripts were tested on following libraries and versions:

  1. pytorch==1.7.1
  2. transformers==4.8.2

Inference: How to use it for parsing

  • Download the code from the repo DMRST_Parser.
  • Put the SegMerge_Infer.py to the repo folder.
  • Put your text paragraph to the file ./data/text_for_inference.txt.
  • Pre-trained model checkpoint should be downloaded and saved at ./depth_mode/Savings/ (see repo DMRST_Parser).
  • Run the script SegMerge_Infer.py to obtain the EDU segmentationg and merge result. See the script for detailed model output.
  • We recommend users to run the parser on a GPU-equipped environment.

Citation

If the work is helpful, please cite our papers in your publications, reports, slides, and docs.

@inproceedings{liu2019exploiting,
  title={Exploiting discourse-level segmentation for extractive summarization},
  author={Liu, Zhengyuan and Chen, Nancy},
  booktitle={Proceedings of the 2nd Workshop on New Frontiers in Summarization},
  pages={116--121},
  year={2019}
}

@inproceedings{liu-etal-2021-dmrst,
    title = "{DMRST}: A Joint Framework for Document-Level Multilingual {RST} Discourse Segmentation and Parsing",
    author = "Liu, Zhengyuan and Shi, Ke and Chen, Nancy",
    booktitle = "Proceedings of the 2nd Workshop on Computational Approaches to Discourse",
    month = nov,
    year = "2021",
    address = "Punta Cana, Dominican Republic and Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.codi-main.15",
    pages = "154--164",
}