Using Images to Find Context-Independent Word Representations in Vector Space

TL;DR

This paper introduces a novel approach to generate context-independent word vectors using images and dictionary meanings, leveraging auto-encoders to achieve competitive performance with faster training times.

Contribution

The paper presents a new method that uses images and dictionary definitions to create word vectors, reducing training time compared to traditional context-based models.

Findings

Performs comparably to context-based methods in word similarity tasks

Requires significantly less training time

Effective in concept categorization and outlier detection

Abstract

Many methods have been proposed to find vector representation for words, but most rely on capturing context from the text to find semantic relationships between these vectors. We propose a novel method of using dictionary meanings and image depictions to find word vectors independent of any context. We use auto-encoder on the word images to find meaningful representations and use them to calculate the word vectors. We finally evaluate our method on word similarity, concept categorization and outlier detection tasks. Our method performs comparably to context-based methods while taking much less training time.