Situated Dialogue Learning through Procedural Environment Generation

TL;DR

This paper introduces a method for training goal-driven agents in textual environments by procedurally generating new worlds and quests, creating a curriculum that improves generalization to unseen tasks.

Contribution

It presents a novel approach of curriculum learning through procedural environment generation in a large-scale text adventure game.

Findings

Procedural generation of worlds enhances agent performance.

Curriculum based on quest rarity improves generalization.

Agents achieve better zero-shot performance on unseen quests.

Abstract

We teach goal-driven agents to interactively act and speak in situated environments by training on generated curriculums. Our agents operate in LIGHT (Urbanek et al. 2019) -- a large-scale crowd-sourced fantasy text adventure game wherein an agent perceives and interacts with the world through textual natural language. Goals in this environment take the form of character-based quests, consisting of personas and motivations. We augment LIGHT by learning to procedurally generate additional novel textual worlds and quests to create a curriculum of steadily increasing difficulty for training agents to achieve such goals. In particular, we measure curriculum difficulty in terms of the rarity of the quest in the original training distribution -- an easier environment is one that is more likely to have been found in the unaugmented dataset. An ablation study shows that this method of learning from the tail of a distribution results in significantly higher generalization abilities as measured by zero-shot performance on never-before-seen quests.