Visual Word2Vec (vis-w2v): Learning Visually Grounded Word Embeddings Using Abstract Scenes

TL;DR

This paper introduces vis-w2v, a model that learns visually grounded word embeddings using abstract scenes, capturing semantic relatedness beyond text and improving performance on several vision-language tasks.

Contribution

The paper presents a novel method for learning visually grounded word embeddings using abstract scenes, enhancing semantic understanding in vision-language applications.

Findings

Embeddings capture fine-grained, visually grounded semantic relatedness.

Improves over text-only embeddings on assertion classification, paraphrasing, and image retrieval.

Uses abstract scenes from clipart for visual grounding.

Abstract

We propose a model to learn visually grounded word embeddings (vis-w2v) to capture visual notions of semantic relatedness. While word embeddings trained using text have been extremely successful, they cannot uncover notions of semantic relatedness implicit in our visual world. For instance, although "eats" and "stares at" seem unrelated in text, they share semantics visually. When people are eating something, they also tend to stare at the food. Grounding diverse relations like "eats" and "stares at" into vision remains challenging, despite recent progress in vision. We note that the visual grounding of words depends on semantics, and not the literal pixels. We thus use abstract scenes created from clipart to provide the visual grounding. We find that the embeddings we learn capture fine-grained, visually grounded notions of semantic relatedness. We show improvements over text-only word embeddings (word2vec) on three tasks: common-sense assertion classification, visual paraphrasing and text-based image retrieval. Our code and datasets are available online.