TransLaw: A Large-Scale Dataset and Multi-Agent Benchmark Simulating Professional Translation of Hong Kong Case Law
Researchers have developed TransLaw, a novel multi-agent framework designed to improve the professional translation of Hong Kong case law. This system integrates a specialized glossary, retrieval-augmented generation, and iterative feedback, building upon the newly released HKCFA Judgment 97-22, a large-scale parallel corpus of translated judgments. Benchmarking against 13 LLMs, TransLaw demonstrated significant improvements over single-agent approaches, though human evaluation indicated it still lags behind human legal translators in stylistic naturalness. AI
IMPACT This research could lead to more accurate and efficient translation of legal documents, potentially impacting legal professionals and international legal frameworks.