2023.10.27
Overview
A training system and a training method for a data model in a specific field are provided. The training method includes configuring a computing device to perform the following steps: generating, by a training set generation module, a corresponding training data set according to knowledge graph in a specific field; updating a data model based on the training data set; generating, by a training set generation module, training input text corresponding to the knowledge graph in the specific field; inputting the training input text into the data model to obtain training output text; scoring, by a scoring module, based on correlation between the training output text and the knowledge graph in the specific filed to generate scores; adjusting, by a reinforcement learning module, parameters of the data model according to the scores and an optimization goal of a reward model until the scores meets a training completion condition, and using the data model as a data model in the specific field.Category
發明
Patented
112141176
發明第I877867號
Filing Date
2023.10.27Expired Date
2043.10.26Notification
2025.11.21