English
version of this page
Extracting Ontologies from Language Models via Queries
Language models such as GPT have been trained on large amounts of data and can generate text in various domains of knowledge. However, these models lack the ability of performing complex logical reasoning. Ontologies, on the other hand, are useful for representing the relevant knowledge of a domain of interest in a format that allows for logical reasoning in an automated way. Though, building ontologies is known to be an expensive and time-consuming task.
In this project, the candidate will explore the possibility of extracting an ontology from a language model by posing queries in Angluin’s exact learning framework. In the exact learning framework, a learner attempts to extract information from a teacher by posing queries (which represent questions). The idea in this project is to see the language model as the teacher and apply an algorithm designed for learning EL ontologies via queries and counterexamples to this setting.
See full description here.
Publisert 7. sep. 2023 08:27
- Sist endret 18. sep. 2023 19:00
