Cross-modality matching based on Fisher Vector with neural word embeddings and deep image features | IEEE Conference Publication | IEEE Xplore