Visual Word2Vec (vis-w2v): Learning Visually Grounded Word Embeddings Using Abstract Scenes
This addresses the challenge of grounding diverse semantic relations into vision for applications in AI and NLP, though it is incremental by building on existing embedding methods.
The paper tackles the problem of learning word embeddings that capture visual semantic relatedness, which text-only embeddings miss, by using abstract scenes from clipart for grounding. It shows improvements over word2vec on tasks like common-sense assertion classification, visual paraphrasing, and text-based image retrieval.
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.