Understanding route directions in human-robot dialogue

Abstract

This paper discusses some of the challenges in building a robot that is supposed to autonomously navigate in an urban environment by asking pedestrians for route directions. We present a novel Wizard-of-Oz setup for collecting route direction dialogues, and a novel approach for datadriven semantic interpretation of route descriptions into route graphs. The results indicate that it is indeed possible to get people to freely describe routes which can be automatically interpreted into a route graph that may be useful for robot navigation. The purpose of the IURO project 1 – a successor of the ACE project (Bauer et al., 2009) – is to explore how robots can be endowed with capabilities for extracting missing information from humans through spoken interaction. The test scenario for the project is to build a robot that can autonomously navigate in a real urban environment and enquire human passers-by for route directions (see Figure 1). 1