A Corpus for Reasoning About Natural Language Grounded in Photographs

Code

• github.com/lil-lab/nlvr
  Officialnone★ 0
• github.com/mcgill-nlp/incontext-code-generation
  pytorch★ 8

Abstract

We introduce a new dataset for joint reasoning about natural language and images, with a focus on semantic diversity, compositionality, and visual reasoning challenges. The data contains 107,292 examples of English sentences paired with web photographs. The task is to determine whether a natural language caption is true about a pair of photographs. We crowdsource the data using sets of visually rich images and a compare-and-contrast task to elicit linguistically diverse language. Qualitative analysis shows the data requires compositional joint reasoning, including about quantities, comparisons, and relations. Evaluation using state-of-the-art visual reasoning methods shows the data presents a strong challenge.

Tasks

Reproductions