What can you do with a rock? Affordance extraction via word embeddings

TL;DR

This paper introduces a method for extracting affordances using word embeddings trained on Wikipedia, enabling reinforcement learning agents to improve decision-making and human-likeness in text-based environments.

Contribution

The paper presents a novel approach to affordance extraction using word embeddings as a knowledge base, enhancing action selection in reinforcement learning agents.

Findings

Affordance-based action selection improves agent performance in most cases.

The method reduces total learning steps despite increased per-step complexity.

Agent's actions become more human-like with affordance integration.

Abstract

Autonomous agents must often detect affordances: the set of behaviors enabled by a situation. Affordance detection is particularly helpful in domains with large action spaces, allowing the agent to prune its search space by avoiding futile behaviors. This paper presents a method for affordance extraction via word embeddings trained on a Wikipedia corpus. The resulting word vectors are treated as a common knowledge database which can be queried using linear algebra. We apply this method to a reinforcement learning agent in a text-only environment and show that affordance-based action selection improves performance most of the time. Our method increases the computational complexity of each learning step but significantly reduces the total number of steps needed. In addition, the agent's action selections begin to resemble those a human would choose.