Dependency-based convolutional neural networks for sentence embedding

In sentence modeling and classification, convolutional neural network approaches have recently achieved state-of-the-art re-sults, but all such efforts process word vec-tors sequentially and neglect long-distance dependencies. To combine deep learn-ing with linguistic structures, we pro-pose a dependency-based convolution ap-proach, making use of tree-based n-grams rather than surface ones, thus utlizing non-local interactions between words. Our model improves sequential baselines on all four sentiment and question classification tasks, and achieves the highest published accuracy on TREC.