ODYSSEY

The project Odyssey has received an Innovate UK grant under the "Professional & Financial Services Data Access Demonstrators: ESG" funding competition. The project consortium includes Tabled Technologies Limited, Legal-Pythia LLP, The National Archives, Swansea University, Keele University, and Northumbria University.

• Enrich The National Archives LegalDocML files with additional metadata that model general rather than contextual normative rules and their interrelations, temporal validity, legal interpretations and arguments, etc. My role in the project mostly concerns the co-design of this metadata model.


• Fine-tune and evaluate Large Language Models (LLMs) on the enriched datasets. State of the art LLMs (GPT, Claude2, Llama, etc.) will be assessed, to hone the accuracy, ethics and legal suitability for use.


• Create plain-English, standardised LLM prompts for legal subject matter information summarisation for consistent input into LLMs.


• For ESG purposes, the LegalPath A2J app previously developed by Tabled Technologies Limited and Northumbria University will then empower litigants in person and SMEs to construct standardized prompts for accessing LLMs fine-tuned on the enriched datasets.