Dynamic and Multi-Match Answer Selection Model for Automobile Question Answering

Abstract

Automobile question answering is mainly integrated into the vehicle system to provide
users with some common query functions about vehicle questions or instructions. One of the key
sub-tasks in automobile question answering system is answer selection. In this paper, we build
different neural network QA models with automobile related QA dataset, and improve the basic
convolutional neural network answer selection model and the recurrent neural network answer
selection model. Finally, we propose to use dynamic word embedding and multiple matching model
to build the dynamic and multi-match answer selection model for automobile question answering.