Looking at Europe’s German-speaking DACH region (i.e. Germany, Austria, Switzerland) with a population of some 101 million, the vast majority of them are non-legal experts. The complexity of legal language and domain-specific terminology may make these legal subjects face obstacles in getting access to the rule of law, to their rights, obligations and responsibilities. PLAInLaw4U is an innovative multi-disciplinary RL4LLM model (Pternea, Singh, Chakraborty, Oruganti, Milletari, Bapat, and Jiang, 2024) under development. It is aimed at closing the accessibility gap for various non-expert target groups including non-domain professionals, non-native German speakers, and law students. It is based on the recently introduced intralingual DEplain model, a scientific LLM for simplifying German texts. Including a legal terminology database, PLAInLaw4U aims at further specialisation in simplifying German legal texts to provide plain German accessible information. To grant high performance results, PLAInLaw4U is designed to include prompt engineering to ensure appropriate legal paraphrasing springing from reward-based reinforcement learning (RL). This approach aims to provide the transferability of RL results and comes with a high scale-up potential. The model is expected to provide a robust and user-friendly AI application for a broad and diverse audience. PLAInLaw4U, a scientific open-access LLM model made in EU and currently under development, shall contribute to democratising access to the rule of law in German language and to foster inclusion of a broad and diverse audience in the DACH region and the EU.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

PLAInLaw4U

  • Petra Schön

摘要

Looking at Europe’s German-speaking DACH region (i.e. Germany, Austria, Switzerland) with a population of some 101 million, the vast majority of them are non-legal experts. The complexity of legal language and domain-specific terminology may make these legal subjects face obstacles in getting access to the rule of law, to their rights, obligations and responsibilities. PLAInLaw4U is an innovative multi-disciplinary RL4LLM model (Pternea, Singh, Chakraborty, Oruganti, Milletari, Bapat, and Jiang, 2024) under development. It is aimed at closing the accessibility gap for various non-expert target groups including non-domain professionals, non-native German speakers, and law students. It is based on the recently introduced intralingual DEplain model, a scientific LLM for simplifying German texts. Including a legal terminology database, PLAInLaw4U aims at further specialisation in simplifying German legal texts to provide plain German accessible information. To grant high performance results, PLAInLaw4U is designed to include prompt engineering to ensure appropriate legal paraphrasing springing from reward-based reinforcement learning (RL). This approach aims to provide the transferability of RL results and comes with a high scale-up potential. The model is expected to provide a robust and user-friendly AI application for a broad and diverse audience. PLAInLaw4U, a scientific open-access LLM model made in EU and currently under development, shall contribute to democratising access to the rule of law in German language and to foster inclusion of a broad and diverse audience in the DACH region and the EU.