Learning Image Embeddings using Convolutional Neural Networks for Improved Multi-Modal Semantics