SpeakNav: Voice-based Route Description Language Understanding for Template Driven Path Search

Download PDF

Abstract

Many navigation applications take natural language speech as input, which avoids typing in words with their hands and decreases the occurrence of traffic accidents. However, most of them fail to understand a user’s free-form description of a route. In addition, they only support input of a specific source and destination, which cannot satisfy users’ diversified search requirements. We propose a SpeakNav framework that enables users to describe intended routes via speech and recommends appropriate routes. Specifically, we propose a novel route template based BERT (RT-BERT) for route description natural language understanding that extracts the information of the intended POI keywords and distances. Then we formalize a template driven path query based on the extracted information. We develop a hybrid label index to compute the network distance between POIs, and a pivot reverse B-tree (PB-tree) index coupled with a branch and bound (BAB) algorithm to improve the query efficiency. Experiments on real datasets and our generated dataset demonstrate the advantages of RT-BERT in terms of accuracy and BAB in terms of efficiency over the baseline algorithms.