Legal consultation reply method and legal field generative large model training method

• Main Authors: WU YAXUAN, REN MENGXING, LIU YINGJIAN, PENG FEI
• Format: Patent
• Language: Chinese; English
• Published: 29.09.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FORADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORYOR FORECASTING PURPOSES; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE,COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTINGPURPOSES, NOT OTHERWISE PROVIDED FOR

The invention discloses a legal consultation reply method and a legal field generative large model training method, and belongs to the technical field of natural language processing. The method comprises the following steps: pre-training a preset generative large model based on unstructured legal text data and general Chinese data to obtain a first fine-tuning large model; performing low-rank matrix increment weight training on the first fine tuning and enlarging model based on an instruction data set pre-constructed for each legal subdivision field to obtain a second fine tuning and enlarging model corresponding to the legal subdivision field; constructing a reward model according to a pre-constructed score sorting data set and a second fine-tuning enlarging model; and performing reinforcement learning training on the second fine-tuning large model by taking the second fine-tuning large model as an intelligent agent and the reward model as an environment to obtain a legal field generative large model. When t