Convolutional Neural Networks for Sentence Classification

Yoon H. Kim (NYU)

Abstract

Convolutional neural networks (CNN) are a particular type of neural network 
architecture originally invented for computer vision. They are the staple of 
modern vision systems and are becoming increasingly popular in natural language 
processing. In the present work we show that a simple CNN with one layer of 
convolution performs remarkably well on phrase/sentence-level classification 
tasks, when the word embeddings are initialized with those from an unsupervised 
neural language model such as word2vec. We also discuss some tricks of the trade 
to more effectively train deep learning systems for NLP.

Bio

Yoon Kim is an MS student at NYU and a data scientist at American International
Group (AIG). He obtained his bachelor's degree in Mathematics and Economics from
Cornell University and a master's degree in Statistics from Columbia University.
His research is in deep learning methods for NLP.