Composing and Embedding the Words-as-Classifiers Model of Grounded Semantics

TL;DR

This paper investigates how the words-as-classifiers model can be incrementally composed and unified with distributional semantics, enhancing grounded lexical semantics through systematic composition methods and empirical evaluation.

Contribution

It introduces new composition strategies for grounded classifiers and unifies them with distributional embeddings, advancing grounded semantic modeling.

Findings

Different classifier types require distinct composition approaches

Unified grounded and distributional representations improve semantic modeling

Empirical results show the effectiveness of proposed composition methods

Abstract

The words-as-classifiers model of grounded lexical semantics learns a semantic fitness score between physical entities and the words that are used to denote those entities. In this paper, we explore how such a model can incrementally perform composition and how the model can be unified with a distributional representation. For the latter, we leverage the classifier coefficients as an embedding. For composition, we leverage the underlying mechanics of three different classifier types (i.e., logistic regression, decision trees, and multi-layer perceptrons) to arrive at a several systematic approaches to composition unique to each classifier including both denotational and connotational methods of composition. We compare these approaches to each other and to prior work in a visual reference resolution task using the refCOCO dataset. Our results demonstrate the need to expand upon existing composition strategies and bring together grounded and distributional representations.