0 citations0 references

Very Deep Convolutional Networks for Text Classification

2017pp. 1107–1116

Citations Over TimeTop 1% of 2017 papers

Alexis Conneau, Holger Schwenk, Loïc Barrault, Yann LeCun

Abstract

The dominant approach for many NLP tasks are recurrent neural networks, in particular LSTMs, and convolutional neural networks. However, these architectures are rather shallow in comparison to the deep convolutional networks which have pushed the state-of-the-art in computer vision. We present a new architecture (VD-CNN) for text processing which operates directly at the character level and uses only small convolutions and pooling operations. We are able to show that the performance of this model increases with the depth: using up to 29 convolutional layers, we report improvements over the state-ofthe-art on several public text classification tasks. To the best of our knowledge, this is the first time that very deep convolutional nets have been applied to text processing.

Related Papers

→ Deep Residual Learning for Image Recognition(2016)216,943 cited
→ Glove: Global Vectors for Word Representation(2014)33,361 cited
→ Convolutional Neural Networks for Sentence Classification(2014)13,600 cited
Very Deep Convolutional Networks for Natural Language Processing.(2016)