DeepOnto: A Python Package for Ontology Engineering with Deep Learning

Tracking #: 3568-4782

Authors: 
Yuan He
Jiaoyan Chen
Hang Dong
Ian Horrocks
Carlo Allocca
Taehun Kim
Brahmananda Sapkota

Responsible editor: 
Eva Blomqvist

Submission type: 
Tool/System Report
Abstract: 
Integrating deep learning techniques, particularly language models (LMs), with knowledge representation techniques like ontologies has raised widespread attention, urging the need of a platform that supports both paradigms. Although packages such as OWL API and Jena offer robust support for basic ontology processing features, they are not designed to process ontologies into various data formats (e.g., text and graph) crucial for supporting subsequent, deep learning-based applications. Moreover, widely-used ontology APIs are primarily Java-based while deep learning frameworks like PyTorch and Tensorflow are mainly for Python programming. To address the needs, we present DeepOnto, a Python package designed for ontology engineering with deep learning. The package encompasses a core ontology processing module founded on the widely-recognised and reliable OWL API, encapsulating its fundamental features in a more “Pythonic” manner and extending its capabilities to incorporate other essential components including reasoning, verbalisation, normalisation, taxonomy, projection, and more. Building on this module, DeepOnto offers a suite of tools, resources, and algorithms that support various ontology engineering tasks, such as ontology alignment and completion, by harnessing deep learning methods, primarily pre-trained LMs. In this paper, we also demonstrate the practical utility of DeepOnto through two use-cases: the Digital Health Coaching in Samsung Research UK and the Bio-ML track of the Ontology Alignment Evaluation Initiative (OAEI).
Full PDF Version: 
Tags: 
Reviewed

Decision/Status: 
Accept

Solicited Reviews:
Click to Expand/Collapse
Review #1
Anonymous submitted on 15/Jan/2024
Suggestion:
Accept
Review Comment:

I would like to thank the authors for addressing my comments.
With the additional changes, I can suggest to accept the paper.

Review #2
By Daniel Garijo submitted on 25/Feb/2024
Suggestion:
Minor Revision
Review Comment:

I thank the authors for going through my comments. While this version addresses most of the comments in my review, I still have minor remarks about the paper. I list them below following the order of the responses.

- Footnote 18 is wrong (it duplicates footnote 19)

- The authors have added: "DeepOnto is actively being integrated into various ongoing projects, details of which remain confidential due to anonymity clauses". This does not answer my question, which was "Is there any evidence about usage of the tool in other tasks or projects besides the ones reported in the paper?" Please be precise with the reporting. If there are projects using the tool, please state how many. If there are none, then state so as well. I do not buy the argument that this information cannot be disclosed, as you are not stating any information about the projects themselves.

- This sentence: "OWL API and Jena offer robust support for basic ontology processing features, they are not designed to process ontologies into various data formats (e.g., text and graph)" does not seem correct. OWL API and Jena do support process ontologies into various RDF formats. RDF represent the information in a graph. The text in the introduction expands on this, but this sentence is misleading.

- Figure 1 still has some pending issues. There are two types of arrows: >> and ->. Do these arrows mean different things? Do colors have a meaning? Also, some arrows are directed, and others are non directed. Please clarify this in the figure, or at least in its caption.

- Response: We have added that the human evaluation results were based on consensus of three domain experts.
- What does this mean exactly? That the three experts had to agree, or that it was based on the majority's vote? Also, what happened if there was disagreement? Was there any Fleis-Kappa agreement calculated on the consensus? If so, why is it not reported?