Daniel Hewlett

Teaching and executing verb phrases

This paper describes a framework for an agent to learn models of verb-phrase meanings from human teachers and combine these models with environmental dynamics to enact verb commands. The framework extends prior work in apprenticeship learning and leverages recent advancements in modeling activities and planning in domains with multiple objects. We show how to both learn a verb model as a relational finite state machine and how to turn this model into reward and heuristic functions that can then be composed with an MDP model of an environment. The resulting “combined model” can then be efficiently searched by a planner to enact a verb command in this environment. Our experiments in simulated robot domains show this framework can be used to quickly teach verb commands and improves over the current state of the art method.

Coarse-to-Fine Question Answering for Long Documents

We present a framework for question answering that can efficiently scale to longer documents while maintaining or even improving performance of state-of-the-art models. While most successful approaches for reading comprehension rely on recurrent neural networks (RNNs), running them over long documents is prohibitively slow because it is difficult to parallelize over sequences. Inspired by how people first skim the document, identify relevant parts, and carefully read these parts to produce an answer, we combine a coarse, fast model for selecting relevant sentences and a more expensive RNN for producing the answer from those sentences. We treat sentence selection as a latent variable trained jointly from the answer only using reinforcement learning. Experiments demonstrate state-of-the-art performance on a challenging subset of the WikiReading dataset and on a new dataset, while speeding up the model by 3.5x-6.7x.