Training an adaptive dialogue policy for interactive learning of visually grounded word meanings
This work addresses the challenge of efficient language learning for AI systems through human interaction, though it is incremental as it builds on existing frameworks like DS-TTR and focuses on simulated environments.
The paper tackles the problem of interactive learning of visually grounded word meanings by developing a multi-modal dialogue system that integrates semantic parsing with visual classifiers, and shows that dialogue policies and capabilities affect learning accuracy and tutor effort, ultimately training an adaptive policy to balance classifier accuracy and tutoring costs.
We present a multi-modal dialogue system for interactive learning of perceptually grounded word meanings from a human tutor. The system integrates an incremental, semantic parsing/generation framework - Dynamic Syntax and Type Theory with Records (DS-TTR) - with a set of visual classifiers that are learned throughout the interaction and which ground the meaning representations that it produces. We use this system in interaction with a simulated human tutor to study the effects of different dialogue policies and capabilities on the accuracy of learned meanings, learning rates, and efforts/costs to the tutor. We show that the overall performance of the learning agent is affected by (1) who takes initiative in the dialogues; (2) the ability to express/use their confidence level about visual attributes; and (3) the ability to process elliptical and incrementally constructed dialogue turns. Ultimately, we train an adaptive dialogue policy which optimises the trade-off between classifier accuracy and tutoring costs.