Learning to interpret natural language commands through human-robot dialog

摘要

Intelligent robots frequently need to understand re-quests from naive users through natural language. Previous approaches either cannot account for lan-guage variation, e.g., keyword search, or require gathering large annotated corpora, which can be ex-pensive and cannot adapt to new variation. We in-troduce a dialog agent for mobile robots that under-stands human instructions through semantic pars-ing, actively resolves ambiguities using a dialog manager, and incrementally learns from human-robot conversations by inducing training data from user paraphrases. Our dialog agent is implemented and tested both on a web interface with hundreds of users via Mechanical Turk and on a mobile robot over several days, tasked with understanding nav-igation and delivery requests through natural lan-guage in an office environment. In both contexts, We observe significant improvements in user satis-faction after learning from conversations. 1