Perspective-Corrected Spatial Referring Expression Generation for Human–Robot Interaction

Intelligent robots designed to interact with humans in real scenarios need be able refer entities actively by natural language. In spatial referring expression generation, the ambiguity is unavoidable due diversity of reference frames, which will lead an understanding gap between and robots. To narrow this gap, paper, we propose a novel perspective-corrected generation (PcSREG) approach for human-robot interaction considering selection frames. The task simplified into process generating diverse relation units. First, pick out all landmarks these units according entropy preference allow its updating through stack model. Then possible expressions are generated different frame strategies. Finally, evaluate every using probabilistic resolution model find best that satisfies both appropriateness effectiveness. We implement proposed on robot system empirical experiments show our can generate more effective practical applications.