NL-2-SPARQL: Ontology-Based Natural Language Querying over 3D Point Cloud Knowledge Graphs
Loading...
Date
2025
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
The Eurographics Association
Abstract
Point clouds are an asset in a wide range of applications, including heritage monitoring and conservation. However, querying capabilities over such 3D datasets, a fundamental task for their practical usability, is often highly challenging. In particular, difficulties arise when it comes to natural language querying, which can be in turn highly useful in lowering the technical barrier in inspecting 3D point clouds. The present paper explores various approaches for natural language querying over point clouds, leaning for the SPARQL-based query system allowed by the 3DOnt framework for point clouds management, recasting the problem as a natural language to SPARQL translation. After reviewing existing methodologies, we propose NL-2- SPARQL, a novel and flexible neuro-symbolic approach that integrates a Large Language Model (LLM) with a Graph Exploration and Query Building tool (GEQB). We then evaluate this method and demonstrate its application within the 3DOnt framework, highlighting its broader applicability to knowledge graphs in general, beyond this specific 3D-oriented context. A video presentation of the 3DOnt framework is available at https://3dom.fbk.eu/projects/3DOnt.
Description
@inproceedings{10.2312:dh.20253280,
booktitle = {Digital Heritage},
editor = {Campana, Stefano and Ferdani, Daniele and Graf, Holger and Guidi, Gabriele and Hegarty, Zackary and Pescarin, Sofia and Remondino, Fabio},
title = {{NL-2-SPARQL: Ontology-Based Natural Language Querying over 3D Point Cloud Knowledge Graphs}},
author = {Codiglione, Matteo and Remondino, Fabio},
year = {2025},
publisher = {The Eurographics Association},
ISBN = {978-3-03868-277-6},
DOI = {10.2312/dh.20253280}
}