Prompting or Fine-tuning? A Comparative Study of Large Language Models for Taxonomy Construction

Set Up

git clone git@github.com:20001LastOrder/Taxonomy-GPT.git
cd Taxonomy-GPT
conda create -n Taxonomy-GPT python=3.9
conda activate Taxonomy-GPT
pip install -r requirements.txt

Guidline

Step 1: Get Original Input datasets

• Wordnet: We retrieve data from Catherin et al
• ACM CCS: We get data in a form of OWL ontologies from ACM Digital Library You can find all the original inputs from /data/original_input.

Step 2: Data Preprocessing

Please check and run: ccs_data_preprocessing.ipynb and wordnet_data_preprocessing.ipynb

Below is the Taxonomies splitting steps for ACM CCS dataset:

Step 3: Train Model

Check settings in config.yml Run train.py

Step 4: Evaluation

Check evaluation.ipynb View utils.py for evaluation algorithm

Step 5: Violation Measures

violation.ipynb