View Proposal

Proposer
Hadj Batatia
Title
Ontology based LLM fine-tuning
Goal
Develop a software framework to fine-tune LLMs using a ontology of the domain.
Description
Ontology-based LLM fine-tuning leverages an ontology – a structured knowledge model of a specific domain – to enhance a large language model's understanding and performance for specialised tasks. This approach aims to overcome the fragmentation of knowledge found in large raw corpora by using the organised, hierarchical structure of an ontology to guide the LLM's learning process. Methods include OntoTune, which uses in-context learning to align the LLM with the ontology, creating task-specific corpora for fine-tuning and improving domain orientation.
Resources
Background
Url
Difficulty Level
Easy
Ethical Approval
None
Number Of Students
0
Supervisor
Hadj Batatia
Keywords
Degrees
Bachelor of Science in Computer Science
Master of Science in Artificial Intelligence
Master of Science in Data Science
BSc Data Sciences
Back to List