Research on Tibetan medicine intelligent question answering system integrating confrontation training and reinforcement learning

Abstract

In this study, a knowledge graph (KG) based Tibetan medicine intelligent question answering (QA) system model was proposed based on an adversarial learning generative network model, in an attempt to alleviate the scarcity of medical resources, promote the heritage and innovation of Tibetan medicine, and ease the shortage of Tibetan medical information. In this model, the simulated answers were generated via adversarial learning, and subsequently the reinforcement learning was applied for feedback-based optimization, with the ultimate aim of enhancing the accuracy rate of this model. Besides, a triple extraction method based on Tibetan features was proposed to construct a KG dialog set. Finally, this model was subjected to an experiment in Chinese and Tibetan datasets, with the results indicating that the accuracy of this intelligent QA model incorporating adversarial networks and reinforcement learning was higher than other models.