Please use this identifier to cite or link to this item: https://hdl.handle.net/10316/115592
Title: Question Answering over Linked Data with GPT-3
Authors: Faria, Bruno
Perdigão, Dylan 
Oliveira, Hugo Gonçalo
Keywords: SPARQL Generation; Prompt Engineering; Few-Shot Learning; Question Answering; GPT-3
Issue Date: 15-Oct-2023
Publisher: Schloss Dagstuhl – Leibniz-Zentrum für Informatik
Project: C645008882-00000055 
metadata.degois.publication.title: Open Access Series in Informatics (OASIcs)
metadata.degois.publication.volume: 113
metadata.degois.publication.location: 12th Symposium on Languages, Applications and Technologies (SLATE 2023)
Abstract: This paper explores GPT-3 for answering natural language questions over Linked Data. Different engines of the model and different approaches are adopted for answering questions in the QALD-9 dataset, namely: zero and few-shot SPARQL generation, as well as fine-tuning in the training portion of the dataset. Answers retrieved by the generated queries and answers generated directly by the model are also compared. Overall results are generally poor, but several insights are provided on using GPT-3 for the proposed task.
URI: https://hdl.handle.net/10316/115592
DOI: 10.4230/OASIcs.SLATE.2023.1
Rights: openAccess
Appears in Collections:FCTUC Eng.Informática - Livros e Capítulos de Livros

Files in This Item:
File Description SizeFormat
OASIcs.SLATE.2023.1.pdf1.63 MBAdobe PDFView/Open
Show full item record

Page view(s)

48
checked on Oct 30, 2024

Download(s)

14
checked on Oct 30, 2024

Google ScholarTM

Check

Altmetric

Altmetric


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.