Language model analysis for ontology subsumption inference
Investigating whether pre-trained language models (LMs) can function as knowledge bases (KBs) has raised wide research interests recently. However, existing works focus on simple, triple-based, relational KBs, but omit more sophisticated, logic-based, conceptualised KBs such as OWL ontologies. To in...
| Главные авторы: | , , , , |
|---|---|
| Формат: | Conference item |
| Язык: | English |
| Опубликовано: |
Association for Computational Linguistics
2023
|
| _version_ | 1826315123926499328 |
|---|---|
| author | He, Y Chen, J Jimenez-Ruiz, E Dong, H Horrocks, I |
| author_facet | He, Y Chen, J Jimenez-Ruiz, E Dong, H Horrocks, I |
| author_sort | He, Y |
| collection | OXFORD |
| description | Investigating whether pre-trained language models (LMs) can function as knowledge bases (KBs) has raised wide research interests recently. However, existing works focus on simple, triple-based, relational KBs, but omit more sophisticated, logic-based, conceptualised KBs such as OWL ontologies. To investigate an LM's knowledge of ontologies, we propose ONTOLAMA, a set of inference-based probing tasks and datasets from ontology subsumption axioms involving both atomic and complex concepts. We conduct extensive experiments on ontologies of different domains and scales, and our results demonstrate that LMs encode relatively less background knowledge of Subsumption Inference (SI) than traditional Natural Language Inference (NLI) but can improve on SI significantly when a small number of samples are given. We will open-source our code and datasets. |
| first_indexed | 2024-12-09T03:18:22Z |
| format | Conference item |
| id | oxford-uuid:5bb7d842-8a26-4d22-bd01-36c16f0e0b44 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-12-09T03:18:22Z |
| publishDate | 2023 |
| publisher | Association for Computational Linguistics |
| record_format | dspace |
| spelling | oxford-uuid:5bb7d842-8a26-4d22-bd01-36c16f0e0b442024-11-05T11:25:50ZLanguage model analysis for ontology subsumption inferenceConference itemhttp://purl.org/coar/resource_type/c_5794uuid:5bb7d842-8a26-4d22-bd01-36c16f0e0b44EnglishSymplectic ElementsAssociation for Computational Linguistics2023He, YChen, JJimenez-Ruiz, EDong, HHorrocks, IInvestigating whether pre-trained language models (LMs) can function as knowledge bases (KBs) has raised wide research interests recently. However, existing works focus on simple, triple-based, relational KBs, but omit more sophisticated, logic-based, conceptualised KBs such as OWL ontologies. To investigate an LM's knowledge of ontologies, we propose ONTOLAMA, a set of inference-based probing tasks and datasets from ontology subsumption axioms involving both atomic and complex concepts. We conduct extensive experiments on ontologies of different domains and scales, and our results demonstrate that LMs encode relatively less background knowledge of Subsumption Inference (SI) than traditional Natural Language Inference (NLI) but can improve on SI significantly when a small number of samples are given. We will open-source our code and datasets. |
| spellingShingle | He, Y Chen, J Jimenez-Ruiz, E Dong, H Horrocks, I Language model analysis for ontology subsumption inference |
| title | Language model analysis for ontology subsumption inference |
| title_full | Language model analysis for ontology subsumption inference |
| title_fullStr | Language model analysis for ontology subsumption inference |
| title_full_unstemmed | Language model analysis for ontology subsumption inference |
| title_short | Language model analysis for ontology subsumption inference |
| title_sort | language model analysis for ontology subsumption inference |
| work_keys_str_mv | AT hey languagemodelanalysisforontologysubsumptioninference AT chenj languagemodelanalysisforontologysubsumptioninference AT jimenezruize languagemodelanalysisforontologysubsumptioninference AT dongh languagemodelanalysisforontologysubsumptioninference AT horrocksi languagemodelanalysisforontologysubsumptioninference |