This repo contains the code and data for the paper at the Findings of ACL 2021:

How Good Is NLP? A Sober Look at NLP Tasks through the Lens of Social Impact [Paper link]

by Zhijing Jin, Geeticka Chauhan, Brian Tse, Mrinmaya Sachan, Rada Mihalcea.

To cite the paper:

@inproceedings{jin2020good,
    title = "How Good Is {NLP}? {A} Sober Look at {NLP} Tasks through the Lens of Social Impact",
    author = "Jin, Zhijing  and
      Chauhan, Geeticka  and
      Tse, Brian  and
      Sachan, Mrinmaya  and
      Mihalcea, Rada",
    booktitle = "Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021",
    month = aug,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/2106.02359",
    doi = "10.18653/v1/2021.findings-acl.273",
    pages = "3099--3113",
}

Visualizations

To check out the visualizations of NLP4SocialGood papers, please see the image files in data/visualization/

Annotated Data

See data/acl_long_clean.csv for our final data.

The format is (paper_name, Stage, track, social good domain, author)

• paper_name: taken from the list of accepted papers at ACL 2020
• Stage: taking values from {1,2,3,4}, classified according to Section 3.1 of our position paper
• track: adapted from the ACL tracks
• social good domain: taking values from {bias mitigation, education, equality, fighting misinformation, green NLP, healthcare, interpretability, legal applications, low-resource language, mental healthcare, robustness, science literature parsing, and others}
• author: taken from the list of accepted papers at ACL 2020

Codes

Preparing for the environment

We use Python 3.

pip install code/requirements.txt

Run the code

python code/visualize_data.py

Pull Requests

We welcome Pull Requests on improving the data csv files, or the codes.