Towards Learning Object Affordance Priors from Technical Texts

TL;DR

This paper explores extracting object affordance priors from technical texts to enhance AI assistants' understanding of entity capabilities and modalities, aiming to improve safety and effectiveness in everyday activities.

Contribution

It introduces a conceptual framework for deriving modality and affordance knowledge from non-figurative texts using grammatical co-occurrence analysis.

Findings

Proposes a method to extract semantic relations from texts

Highlights potential for improving AI common sense knowledge

Discusses limitations and future framework adoption

Abstract

Everyday activities performed by artificial assistants can potentially be executed naively and dangerously given their lack of common sense knowledge. This paper presents conceptual work towards obtaining prior knowledge on the usual modality (passive or active) of any given entity, and their affordance estimates, by extracting high-confidence ability modality semantic relations (X can Y relationship) from non-figurative texts, by analyzing co-occurrence of grammatical instances of subjects and verbs, and verbs and objects. The discussion includes an outline of the concept, potential and limitations, and possible feature and learning framework adoption.