Search papers, labs, and topics across Lattice.
This paper introduces GreekBarRetrieval, a benchmark designed to enhance statutory retrieval for Greek legal question answering, addressing a significant gap in the field. The benchmark includes 283 bar-exam questions paired with relevant case facts and a pool of 6,308 statutory articles, requiring effective mapping from everyday language to formal legal terminology. Key findings reveal that while dense retrieval methods outperform sparse ones, integrating LLM-based query reformulation significantly boosts retrieval performance across all tested methods, achieving the highest scores in Recall@100, nDCG, and MAP.
LLM-based query reformulation not only bridges the gap in retrieval performance but also sets a new standard for legal question answering in Greek statutory contexts.
Statutory retrieval is necessary for citation-grounded legal question answering, but remains underexplored for Greek. We introduce GreekBarRetrieval, a public retrieval benchmark derived from, and complementing GreekBarBench, which did not include retrieval. The new benchmark comprises 283 bar-exam questions, each accompanied by the facts of the case it refers to, and 6,308 candidate statutory articles to retrieve from. Questions and facts are stated in everyday language, but need to be mapped to the formal terminology of statutes and their abstract legal concepts. A further complication is that not all of the case facts are relevant to each question of a case. Experimenting with three BM25 variants and nine dense retrievers, we find that vanilla dense retrieval far outperforms vanilla sparse retrieval in Recall@100. However, LLM-based query reformulation helps BM25 close that gap, while also improving dense retrieval. With a ten-round ReAct-like LLM reformulation loop that we introduce, BM25 improves further in Recall@100 and obtains the best nDCG and MAP scores of all tested retrievers. Query reformulation also outperforms pseudo-relevance feedback, sparse-dense fusion, and English translation.
