Training Data and Transfer Learning in Legal Argument Mining

Tue, 13 Oct 2026 00:00:00 +0000

Context

Legal argument mining identifies, extracts, and reconstructs the arguments of judicial texts automatically. It can reduce manual analysis and support legal research, document retrieval, summarization, and outcome prediction. Transfer learning trains a model on one dataset and applies it to another.

Objective

This proposal studies large language models and traditional models, and the impact of the available data in quantity and in domain. The data can serve as training data, as demonstrations in the prompt, or as documents that retrieval-augmented generation (RAG) finds for the model. The proposal also measures transfer learning between two legal domains: fiscal state aid and equal treatment.

The two questions of the proposal, built from its description: how the quantity and the form of the data change legal argument mining, and how a model transfers between two legal domains.

Transfer Learning | Language Technologies Lab

Training Data and Transfer Learning in Legal Argument Mining

Context

Objective