aiqbench analysis
This research demonstrates that training neural language models with visual context significantly improves next-word prediction performance and generalization to text-only tasks. The findings support the situated cognition theory, showing that multi-modal learning environments yield more robust AI systems across various architectures and languages.
Like a Baby: Visually Situated Neural Language Acquisition
arXiv:1805.11546v3 Announce Type: replace-cross 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.
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.
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
arXiv:1805.11546 [cs.CL]
(or arXiv:1805.11546v3 [cs.CL] for this version)
Full text under the publisher's open licence · 146 words
