SpeakNav

Abstract

Many navigation applications take natural language speech as input, which avoids users typing in words and thus improves traffic safety. However, navigation applications often fail to understand a user's free-form description of a route. In addition, they only support input of a specific source or destination, which does not enable users to specify additional route requirements. We propose a SpeakNav framework that enables users to describe intended routes via speech and then recommends appropriate routes. Specifically, we propose a novel Route Template based Bidirectional Encoder Representation from Transformers (RT-BERT) model that supports the understanding of natural language route descriptions. The model enables extraction of information of intended POI keywords and related distances. Then we formalize a template-driven path query that uses the extracted information. To enable efficient query processing, we develop a hybrid label index for computing network distances between POIs, and we propose a branch-and-bound algorithm along with a pivot reverse B-tree (PB-tree) index. Experiments with real and synthetic data indicate that RT-BERT offers high accuracy and that the proposed algorithm is capable of outperforming baseline algorithms.

Distributed Processing of k Shortest Path Queries over Dynamic Road Networks

“…4a, (1,8), and (2, 3), meaning there are 3 vfrags of unit weight 1/3, 4 vfrags of unit weight 1/2, and so on. Thus, BD(P ′ 1 (13,14)) can be computed using the 8 smallest unit weights, consisting of 3 unit weights of 1/3, 4 unit weights of 1/2, and 1 unit weight of 1, with the result being 4.…”

“…Identifying KSPs in a dynamic road network is an essential building block in many location-based services such as route navigation and ride-sharing [1], [2], [3], [4]. Most of the existing work to address the KSP problem in a road network (or more generally, in a graph) assumes a centralized approach [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], and is incapable of handling large volumes of concurrent queries over a dynamic graph for two main reasons. First, the query processing strategies employed are sequential, namely they generate the (i+1) th shortest path based on the i th shortest path, which limits their scalability with respect to the number of concurrent queries.…”

Urban Region Embedding via Multi-View Contrastive Prediction

“…house price prediction, and population density inference (Liu et al 2021;Li et al 2022;Liu et al 2023;Huang et al 2023;Xu et al 2023b;Li et al 2023). This trend can also be attributed to the prosperity of mobile sensing technologies, which has led to the rapid accumulation of urban sensing data, such as human trajectories or points-of-interest (POIs) (Zheng et al 2020(Zheng et al , 2021Chen, Yu, and Liu 2018;Zhang, Zhao, and Chen 2022; Figure 1: Illustration of (a) multi-view fusion paradigm and our proposed (b) consistency learning paradigm for region embedding. In the right figure, the solid and dotted rectangles denote the region representations Z a and Z m from the attribute and mobility views, respectively.…”