Learning semantic representations using convolutional neural networks for web search

This paper presents a series of new latent semantic models based on a convolutional neural network to learn low-dimensional semantic vectors for search queries and Web documents that significantly outperforms other se-mantic models in retrieval performance.