amphora/korfin-asc

Dataset Card for KorFin-ABSA

Dataset Description

Dataset Summary

The KorFin-ASC is an extension of KorFin-ABSA including 8818 samples with (aspect, polarity) pairs annotated. The samples were collected from KLUE-TC and analyst reports from Naver Finance. Annotation of the dataset is described in the paper Removing Non-Stationary Knowledge From Pre-Trained Language Models for Entity-Level Sentiment Classification in Finance.

Supported Tasks and Leaderboards

This dataset supports the following tasks:

• Aspect-Based Sentiment Classification

Languages

Korean

Dataset Structure

Data Instances

Each instance consists of a single sentence, aspect, and corresponding polarity (POSITIVE/NEGATIVE/NEUTRAL).

{
  "title": "LGU＋ 1분기 영업익 1천706억원…마케팅 비용 감소",
  "aspect": "LG U+",
  'sentiment': 'NEUTRAL', 
  'url': 'https://news.naver.com/main/read.nhn?mode=LS2D&mid=shm&sid1=105&sid2=227&oid=001&aid=0008363739', 
  'annotator_id': 'A_01', 
  'Type': 'single'
}

Data Fields

• title:
• aspect:
• sentiment:
• url:
• annotator_id:
• url:

Data Splits

The dataset currently does not contain standard data splits.

Additional Information

You can download the data via:

from datasets import load_dataset
dataset = load_dataset("amphora/KorFin-ASC")

Please find more information about the code and how the data was collected in the paper Removing Non-Stationary Knowledge From Pre-Trained Language Models for Entity-Level Sentiment Classification in Finance. The best-performing model on this dataset can be found at link.

Licensing Information

KorFin-ASC is licensed under the terms of the cc-by-sa-4.0

Citation Information

Please cite this data using:

@article{son2023removing,
  title={Removing Non-Stationary Knowledge From Pre-Trained Language Models for Entity-Level Sentiment Classification in Finance},
  author={Son, Guijin and Lee, Hanwool and Kang, Nahyeon and Hahm, Moonjeong},
  journal={arXiv preprint arXiv:2301.03136},
  year={2023}
}

Contributions

Thanks to @Albertmade, @amphora for making this dataset.