Now that we have learned core computational linguistics concepts and trained relations from the provided dataset, we can use this learning to implement a language model which can perform a task. 

In this section, we will build a text classification model to perform sentiment analysis. For classification, we will be using a combination of CNN and a pre-trained word2vec model which we learned in the previous section of this chapter. 

We are inspired by Denny Britz's (https://twitter.com/dennybritz) work on Implementing a CNN for Text Classification in TensorFlow (http://www.wildml.com/2015/12/implementing-a-cnn-for-text-classification-in-tensorflow/) in our own CNN and text classification build.  We invite you to review the blog he created to gain a more complete understanding of the internal mechanisms that make CNN's useful for text classification.

As an overview, this architecture starts with an input embedding step, then a 2D convolution utilizing max pooling with multiple filters and softmax activation layer producing the output.