Grounded Language Acquisition From Object and Action Imagery

TL;DR

This paper investigates how emergent language models trained on visual data can develop grounded symbols for object and action recognition, using referential and contrastive learning in visual tasks.

Contribution

It introduces a dual training approach for emergent language models in visual recognition, combining referential and contrastive learning with interpretability methods.

Findings

Symbols can be grounded in visual features using the proposed methods.

Gradient-based interpretability reveals semantic regions associated with symbols.

Embeddings show meaningful clustering related to object and action classes.

Abstract

Deep learning approaches to natural language processing have made great strides in recent years. While these models produce symbols that convey vast amounts of diverse knowledge, it is unclear how such symbols are grounded in data from the world. In this paper, we explore the development of a private language for visual data representation by training emergent language (EL) encoders/decoders in both i) a traditional referential game environment and ii) a contrastive learning environment utilizing a within-class matching training paradigm. An additional classification layer utilizing neural machine translation and random forest classification was used to transform symbolic representations (sequences of integer symbols) to class labels. These methods were applied in two experiments focusing on object recognition and action recognition. For object recognition, a set of sketches produced by human participants from real imagery was used (Sketchy dataset) and for action recognition, 2D trajectories were generated from 3D motion capture systems (MOVI dataset). In order to interpret the symbols produced for data in each experiment, gradient-weighted class activation mapping (Grad-CAM) methods were used to identify pixel regions indicating semantic features which contribute evidence towards symbols in learned languages. Additionally, a t-distributed stochastic neighbor embedding (t-SNE) method was used to investigate embeddings learned by CNN feature extractors.