Few-Example-Event-Extraction

The repo for "Learning Event Extraction From a Few Guideline Examples"

Abstract - Existing fully supervised event extraction models achieve advanced performance with large-scale labeled data. However, when new event types emerge and annotations are scarce, it is hard for the supervised models to master the new types with limited annotations. In contrast, humans can learnto understand new event types with only a few examples in the event extraction guideline. In this paper, we work on a challenging yet more realistic setting, the few-example event extraction. It requires models to learn event extraction with only a few sentences in guidelines as training data, so that we do not need to collect large-scale annotations each time when new event types emerge. As models tend to overfit when trained with only a few examples, we propose knowledge-guided data augmentation to generate valid and diverse sentences from the guideline examples. To help models better leverage the augmented data, we add a consistency regularization to guarantee consistent representations between the augmented sentences and the original ones. Experiments on the standard benchmark ACE-2005 indicate that our method can extract event triggers and arguments effectively with only a few guideline examples.

Data

ACE-Guideline: We collect 194 sentences from Chapters 5 and 6 of the ACE event annotation guideline as a few guideline examples to train models.

• Training split: data/ACE-guideline.train.json

• Development split: data/ACE-guideline.dev.json

The data format are in the OneIE data format.

Code: Knowledge-Guided Data Augmentation

1. Run pip install -r code/requirements.txt

2. Download ProBase and set probase_path in code/KGDA.py

3. Run the knowledge-guided data augmentation:

python code/KGDA.py
--src_files [path to ACE-guideline data]  \
--target_file [dir of save augmented ACE-guideline data] 

Citation

@article{DBLP:journals/taslp/HongZYZ22,
    author    = {Ruixin Hong and
                Hongming Zhang and
                Xintong Yu and
                Changshui Zhang},
    title     = {Learning Event Extraction From a Few Guideline Examples},
    journal   = {{IEEE} {ACM} Trans. Audio Speech Lang. Process.},
    volume    = {30},
    pages     = {2955--2967},
    year      = {2022},
    url       = {https://doi.org/10.1109/TASLP.2022.3202123},
    doi       = {10.1109/TASLP.2022.3202123},
    timestamp = {Tue, 18 Oct 2022 22:17:28 +0200},
    biburl    = {https://dblp.org/rec/journals/taslp/HongZYZ22.bib},
    bibsource = {dblp computer science bibliography, https://dblp.org}
}