A Transformer-based Attention Flow Model for Intelligent Question and Answering Chatbot
Yunze Xiao · 2022
The question and answer system is a key step and core component in building an intelligent voice assistant, and traditional question and answer systems use various deep learning networks to model this task, in our task, we propose a novel neural network-based model structure that uses the Transformer layer to enhance the model semantic feature extraction capability, and to compensate for the fact that the Transformer is not good at capturing location information, we use a bidirectional LSTM layer to enhance the model's temporal feature extraction capability. In the experiments with SQUAD data, our model achieves an EM index of 68.5 and an Fl score of 77.7, which proves the effectiveness of our model.