This repository contains our work for the final project of the course Semantic Web Technology of the MA Information Science program, taught at the University of Groningen.
The goal of our application is to translate RDF (Resource Description Framework) triples to natural language. The Triple Translator (also known as TRIPTRAN) takes a single RDF triple as input and translates it to English.
TRIPTRAN relies on several Python packages - including NLTK - for its functionality. You can easily install these package by running the following command:
pip install -r requirements.txt
After you have succesfully installed all the packes, you first need to
download the data from the WebNLG repository. You can use the
download()
function in datamanager.py
to achieve this. Simply run
the following lines of code in your Python console:
>>> from datamanager import download
>>> download()
The data is stored in XML files in a newly created directory named data.
Once you have succesfully downloaded all the data, you can start using
TRIPTRAN! You can start by running main.py
, which trains the TRIPTRAN
model on the training data of WebNLG 2020. The script then uses WebNLG
test data from 2017 to evaluate the translation performances. Simply run
the following line of code in your console:
py -m main