Skip to content

README.md

Latest commit

 

History

History
64 lines (42 loc) · 2.45 KB

README.md

File metadata and controls

64 lines (42 loc) · 2.45 KB

ECDaily: A Large-scale Benchmark for Emotion Cause Extraction in Conversations

This repository contains the data and code for the paper "ECDaily: A Large-scale Benchmark for Emotion Cause Extraction in Conversations".

Overview

ECDaily is a large-scale benchmark dataset designed for research on emotion cause extraction in conversational contexts. This repository provides the complete dataset along with baseline implementations.

Repository Structure

  • dataset/: Contains the ECDaily dataset split into three files:

    • train.txt: Training set
    • valid.txt: Validation set
    • test.txt: Test set

  • ECE-T5/: Source code for the T5-based baseline model for Emotion Cause Extraction (ECE)

  • ECE-roberta/: Source code for the RoBERTa-based baseline model for Emotion Cause Extraction (ECE)

  • ECPE-T5/: Source code for the T5-based baseline model for Emotion-Cause Pair Extraction (ECPE)

  • ECPE-roberta/: Source code for the RoBERTa-based baseline model for Emotion-Cause Pair Extraction (ECPE)

Dataset

The ECDaily dataset is provided in the dataset/ directory with train, validation, and test splits.

Baselines

We provide four baseline implementations:

  1. ECE-T5: T5-based model for emotion cause extraction
  2. ECE-RoBERTa: RoBERTa-based model for emotion cause extraction
  3. ECPE-T5: T5-based model for emotion-cause pair extraction
  4. ECPE-RoBERTa: RoBERTa-based model for emotion-cause pair extraction

Each baseline folder contains the complete source code and necessary files to reproduce the results.

Citation

If you use this dataset or code in your research, please cite our paper:

@ARTICLE{2025ITAfC..16.1570S,
       author = {{Shen}, Xiangqing and {Li}, Ke and {An}, Jiaming and {Ding}, Zixiang and {Xia}, Rui},
        title = "{ECDaily: A Large-scale Benchmark for Emotion Cause Extraction in Conversations}",
      journal = {IEEE Transactions on Affective Computing},
     keywords = {Emotion cause analysis, emotion cause extraction, emotion recognition in conversations, textual emotion analysis},
         year = 2025,
        month = jan,
       volume = {16},
       number = {3},
        pages = {1570-1580},
          doi = {10.1109/TAFFC.2024.3524124},
       adsurl = {https://ui.adsabs.harvard.edu/abs/2025ITAfC..16.1570S},
      adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}

License

This project is licensed under the terms specified in the LICENSE file.