license: cc-by-4.0
Each line includes one example, represented as a JSON object. The critical fields are:
sentence: The natural language sentence describing the pair of images for this example.
left_url: The URL of the left image in the pair.
right_url: The URL of the right image in the pair.
label: The label: true or false.
identifier: The unique identifier for the image, in the format: split-set_id-pair_id-sentence-id. split is the split of the data (train, test, or development). set_id is the unique identifier of the original eight-image set used in the sentence-writing task. pair_id indicates which of the pairs in the set it corresponds to (and is between 0 and 3). sentence-id indicates which of the sentences is associated with this pair (and is either 0 or 1 -- each image pair is associated with at most two sentences).