Dataset to be considered: ParagraphRetrievalECHR

[JulienGaumez]

Legal professionals often grapple with navigating lengthy legal judgements to pinpoint information that directly address their queries. This paper focus on this task of extracting relevant paragraphs from legal judgements based on the query. We construct a specialized dataset for this task from the European Court of Human Rights (ECtHR) using the case law guides. We eventually end up with 4109 query-judgement pairs with 708 unique queries.

Dataset: https://github.com/TUMLegalTech/ParagraphRetrievalECHR/

[JoelNiklaus]

This is an IR task. We need to find a way of formulating IR tasks so it makes sense for instruction tuning to use this.