Research on machine reading comprehension based on pre-trained model
Guanlin Chen, Rutao Yao, Haiwei Zhou, Tian Li, Wujian Yang · International Journal of Reasoning-based Intelligent Systems · 2022
In order to improve the machine reading ability of the model, this article starts from the high-level semantic information of the text and the concepts of the distillation model. Among the high-level semantic information of the text, part-of-speech information and named entity information are selected as additional information that the model can obtain. Combined with part-of-speech tagging technology and named entity recognition technology, a BERT-HSI model based on a pre-training model and fusion of high-level semantic information is proposed. Later, on the basis of BERT-HSI, the paper started from the perspective of model optimisation, with the concepts in the distillation model as the core, and proposed a capability learning method. Finally, the research method about machine reading comprehension based on the pre-training model is presented, which not only integrates high-level semantic information but also includes the capacity learning process.