Official Repository for paper Ontology-Free General-Domain Knowledge Graph-to-Text Generation Dataset Synthesis using Large Language Model
To use this repository, you must download the provided QuestEval directory. This directory contains modifications to the official QuestEval repository. These modifications are necessary for the specific analyses performed in this project, so please ensure you use the provided version.
git clone https://github.com/daehuikim/WikiOFGraph.git
cd WikiOFGraph
pip install transformers, vllm, openai, datasets
cd qeval
pip install -e .The data used in this project is provided via the Huggingface datasets. You can download and prepare the dataset by running the following:
from datasets import load_dataset
dataset = load_dataset("andreaKIM/WikiOFGraph")
or
you can manually download files from this link.
This repository includes code to generate the WikiOFGraph, as described in the paper.
The process involves several steps, such as data pre-processing, graph extraction, and Data-QuestEval Filtering.
Detailed implementations are provided in the process.
This directory contains the details of qualitative analysis described in the paper.
Detailed analysis scripts and example outputs are provided in the qualitativeAnalysis.
Detailed codes related to experiments are in the experiments.
Trainner and inference codes are implemented using transformers.