Like a Baby: Visually Situated Neural Language Acquisition

TL;DR

This paper demonstrates that neural language models trained with visual context outperform language-only models, supporting the idea that language learning benefits from multi-modal, situated environments similar to how babies learn.

Contribution

Introduces a multi-modal neural architecture that leverages visual context to improve language modeling and shows its effectiveness across multiple languages and model types.

Findings

Multi-modal models reduce perplexity by 2% compared to language-only models.

Fine-tuning BERT embeddings yields a 3.5% improvement.

Visual context benefits persist across languages and model architectures.

Abstract

We examine the benefits of visual context in training neural language models to perform next-word prediction. A multi-modal neural architecture is introduced that outperform its equivalent trained on language alone with a 2\% decrease in perplexity, even when no visual context is available at test. Fine-tuning the embeddings of a pre-trained state-of-the-art bidirectional language model (BERT) in the language modeling framework yields a 3.5\% improvement. The advantage for training with visual context when testing without is robust across different languages (English, German and Spanish) and different models (GRU, LSTM, $\Delta$-RNN, as well as those that use BERT embeddings). Thus, language models perform better when they learn like a baby, i.e, in a multi-modal environment. This finding is compatible with the theory of situated cognition: language is inseparable from its physical context.