Repository files navigation

News 📰

• Refinement todeeponto.onto.normalisation (v0.9.3).
• Refinement todeeponto.onto.taxonomy. (v0.9.2)
• Hot fix to theopenprompt issue by moving it to optional dependencies. (v0.9.1)
• Minor feature enhancement; reorganise package layout. (v0.8.9)
• Deploydeeponto.onto.taxonomy; add the structural reasoner type. (v0.8.8)
• Deploy various new ontology processing functions especially for reasoning and verbalisation; update OAEI utitlities for evaluation. (v0.8.7)
• Minor modifications of certain methods and set all utility methods to direct import. (v0.8.5)
• Deploy OAEI utilities atdeeponto.align.oaei for scripts at the sub-repositoryOAEI-Bio-ML as well as bug fixing. (v0.8.4)
• Bug fixing for BERTMap (stuck at reasoning) and ontology alignment evaluation. (v0.8.3)
• Deploydeeponto.onto.OntologyNormaliser anddeeponto.onto.OntologyProjector (v0.8.0).
• Upload Java dependencies directly and remove mowl from pip dependencies (v0.7.5).
• Deploy thedeeponto.subs.bertsubs anddeeponto.onto.pruning modules (v0.7.0).
• Deploy thedeeponto.probe.ontolama anddeeponto.onto.verbalisation modules (v0.6.0).
• Rebuild the whole package based on the OWLAPI; remove owlready2 from the essential dependencies (fromv0.5.x).

Check the completechangelog andFAQs. The FAQs page does not contain much information now but will be updated according to feedback.

$\textsf{DeepOnto}$ aims to provide building blocks for implementing deep learning models, constructing resources, and conducting evaluation for various ontology engineering purposes.

• Documentation:https://krr-oxford.github.io/DeepOnto/.
• Github Repository:https://github.com/KRR-Oxford/DeepOnto.
• PyPI:https://pypi.org/project/deeponto/.

$\textsf{DeepOnto}$ relies onOWLAPI version 4.5.22 (written in Java) for ontologies.

We follow what has been implemented inmOWL that usesJPype to bridge Python and Java Virtual Machine (JVM). Please check JPype'sinstallation page for successful JVM initialisation.

$\textsf{DeepOnto}$ relies onPytorch for deep learning framework.

We recommend installing Pytorch prior to installing$\textsf{DeepOnto}$ following the commands listed on thePytorch webpage. Notice that users can choose either GPU (with CUDA) or CPU version of Pytorch.

In case the most recent Pytorch version causes any incompatibility issues, use the following command (withCUDA 11.6) known to work:

pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu116

Basic usage of$\textsf{DeepOnto}$ does not rely on GPUs, but for efficient deep learning model training, please make suretorch.cuda.is_available() returnsTrue.

Other dependencies are specified insetup.cfg andrequirements.txt which are supposed to be installed along withdeeponto.

# requiring Python>=3.8pip install deeponto

We have been informed thatopenprompt has a conflict with several other packages that can be hardly addressed on MacOS with M1, so we now set it as an optional dependency. However, it is main dependency of the OntoLAMA code atdeeponto.complete.ontolama. To use OntoLAMA, please installopenprompt separately, or use the following command to install$\textsf{DeepOnto}$:

pip install deeponto[ontolama]

To install the latest, probably unreleased version of deeponto, you can directly install from the repository.

pip install git+https://github.com/KRR-Oxford/DeepOnto.git

Figure: Illustration of DeepOnto's architecture.

The base class of$\textsf{DeepOnto}$ is [Ontology][deeponto.onto.Ontology], which serves as the main entry point for introducing the OWLAPI's features, such as accessing ontology entities, querying for ancestor/descendent (and parent/child) concepts, deleting entities, modifying axioms, and retrieving annotations. See quick usage atload an ontology. Along with these basic functionalities, several essential sub-modules are built to enhance the core module, including the following:

• Ontology Reasoning ([OntologyReasoner][deeponto.onto.OntologyReasoner]): Each instance of$\textsf{DeepOnto}$ has a reasoner as its attribute. It is used for conducting reasoning activities, such as obtaining inferred subsumers and subsumees, as well as checking entailment and consistency.

• Ontology Pruning ([OntologyPruner][deeponto.onto.OntologyPruner]): This sub-module aims to incorporate pruning algorithms for extracting a sub-ontology from an input ontology. We currently implement the one proposed in [2], which introduces subsumption axioms between the asserted (atomic or complex) parents and children of the class targeted for removal.

• Ontology Verbalisation ([OntologyVerbaliser][deeponto.onto.OntologyVerbaliser]): The recursive concept verbaliser proposed in [4] is implemented here, which can automatically transform a complex logical expression into a textual sentence based on entity names or labels available in the ontology. Seeverbalising ontology concepts.

• Ontology Projection ([OntologyProjector][deeponto.onto.OntologyProjector]): The projection algorithm adopted in the OWL2Vec* ontology embeddings is implemented here, which is to transform an ontology's TBox into a set of RDF triples. The relevant code is modified from the mOWL library.

• Ontology Normalisation ([OntologyNormaliser][deeponto.onto.OntologyNormaliser]): The implemented$\mathcal{EL}$ normalisation is also modified from the mOWL library, which is used to transform TBox axioms into normalised forms to support, e.g., geometric ontology embeddings.

• Ontology Taxonomy ([OntologyTaxonomy][deeponto.onto.OntologyTaxonomy]): The taxonomy extracted from an ontology is a directed acyclic graph for the subsumption hierarchy, which is often used to support graph-based deep learning applications.

Individual tools and resources are implemented based on the core ontology processing module. Currently,$\textsf{DeepOnto}$ supports the following:

• BERTMap [1] is a BERT-based ontology matching (OM) system originally developed inrepo but is now maintained in$\textsf{DeepOnto}$. SeeOntology Matching with BERTMap & BERTMapLt.

• Bio-ML [2] is an OM resource that has been used in theBio-ML track of the OAEI. SeeBio-ML: A Comprehensive Documentation.

• BERTSubs [3] is a system for ontology subsumption prediction. We have transformed its originalexperimental code into this project. SeeSubsumption Inference with BERTSubs.

• OntoLAMA [4] is an evaluation of language model for ontology subsumption inference. SeeOntoLAMA: Dataset Overview & Usage Guide for the use of the datasets and the prompt-based probing approach.

• HiT (External) [6] is a hierarchy embedding model derived from re-training BERT-like models in hyperbolic space. SeeHiT Models on Huggingface Hub for options and usage.

!!! license "License"

Copyright 2021-2023 Yuan He.Copyright 2023 Yuan He, Jiaoyan Chen.All rights reserved.Licensed under the Apache License, Version 2.0 (the "License");you may not use this file except in compliance with the License.You may obtain a copy of the License at *<http://www.apache.org/licenses/LICENSE-2.0>*Unless required by applicable law or agreed to in writing, softwaredistributed under the License is distributed on an "AS IS" BASIS,WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.See the License for the specific language governing permissions andlimitations under the License.

Our system papaer for$\textsf{DeepOnto}$ is available atarxiv andios press.

Yuan He, Jiaoyan Chen, Hang Dong, Ian Horrocks, Carlo Allocca, Taehun Kim, and Brahmananda Sapkota.DeepOnto: A Python Package for Ontology Engineering with Deep Learning. Semantic Web, vol. 15, no. 5, pp. 1991-2004, 2024.

!!! credit "Citation"

```@article{he2024deeponto,  author = {He, Yuan and Chen, Jiaoyan and Dong, Hang and Horrocks, Ian and Allocca, Carlo and Kim, Taehun and Sapkota, Brahmananda},  journal = {Semantic Web},  number = {5},  pages = {1991--2004},  title = {DeepOnto: A Python package for ontology engineering with deep learning},  volume = {15},  year = {2024}}```

• [1]Yuan He‚ Jiaoyan Chen‚ Denvar Antonyrajah and Ian Horrocks.BERTMap: A BERT−Based Ontology Alignment System. In Proceedings of 36th AAAI Conference on Artificial Intelligence (AAAI-2022). /arxiv/ /aaai/
• [2]Yuan He‚ Jiaoyan Chen‚ Hang Dong, Ernesto Jiménez-Ruiz, Ali Hadian and Ian Horrocks.Machine Learning-Friendly Biomedical Datasets for Equivalence and Subsumption Ontology Matching. The 21st International Semantic Web Conference (ISWC-2022,Best Resource Paper Candidate). /arxiv/ /iswc/
• [3]Jiaoyan Chen, Yuan He, Yuxia Geng, Ernesto Jiménez-Ruiz, Hang Dong and Ian Horrocks.Contextual Semantic Embeddings for Ontology Subsumption Prediction. World Wide Web Journal （WWWJ-2023). /arxiv/ /wwwj/
• [4]Yuan He‚ Jiaoyan Chen, Ernesto Jiménez-Ruiz, Hang Dong and Ian Horrocks.Language Model Analysis for Ontology Subsumption Inference. Findings of the Association for Computational Linguistics (ACL-2023). /arxiv/ /acl/
• [5]Yuan He, Jiaoyan Chen, Hang Dong, and Ian Horrocks.Exploring Large Language Models for Ontology Alignment. The 22nd International Semantic Web Conference (ISWC-2023 Posters & Demos). /arxiv/ /iswc/
• [6]Yuan He, Zhangdie Yuan, Jiaoyan Chen, and Ian Horrocks.Language Models as Hierarchy Encoders. Advances in Neural Information Processing Systems (NeurIPS 2024). /arxiv/ /[neurips](to appear)/

Please report any bugs or queries byraising a GitHub issue or sending emails to the maintainers (Yuan He or Jiaoyan Chen) through:

first_name.last_name@cs.ox.ac.uk