SentenceTensor

AbstractWe propose to learn distributed sentence representation using text’s visual features as input. Different from the existing methods that render the words or characters of a sentence into images separately, …...