RTG4TE

Source code for the paper: Retrieval-enhanced Template Generation for Template Extraction (NLPCC 2024)

Overview

An example of template extraction. A generic template is extracted for document-level REE task. Two event templates including an `Attack` event template and a `Bombing` event template are extracted for TF task.

Requirements

• python==3.7.16
• transformers==4.6.0
• pandas==1.2.3
• omegaconf==2.0.6
• torch==1.7.0
• sentence-transformers==2.2.0
• sentencepiece==0.1.95
• scipy==1.6.1
• spacy==3.0.0
• spacy-legacy==3.0.12
• nltk==3.5

Data

For TF task, we downloaded the original dataset from GTT. The extracted train, dev, and test files are located in data/tf/. These original data are transformed into our internal format using convert_tf.py.

python convert_tf.py --input_path data/train.json --output_path data/tf_train.json

AS for REE task, we downloaded the original dataset from GRIT. The extracted train, dev, and test files are located in data/ree/. These original data are transformed into our internal format using convert_grit.py.

python convert_grit.py --input_path data/grit_train.json --output_path data/ree_train.json

Usage

Template Filling

python train.py -c config/tf_generative_model.json

Role-filler entity extraction

python train.py -c config/ree_generative_model.json

Acknowledgement

We refer to the code of TempGen. Thanks for their contributions.

Citation