How to Build a Neural Network for NLP Tasks with PyTorch and GPU
A framework for modeling text data using Google Colab
This article is meant to serve as a framework to solve natural language processing (NLP) problems using Python, neural networks, and GPU. Though a high level of math is required to understand everything herein, I wrote this article with the intent that an absolute beginner or seasoned veteran would get something out of it. I would hope a beginner takes away the general framework of solving NLP problems whereas a veteran may use it as a reference for pushing models to GPU in PyTorch.
Let’s get started!
NLP: Data and Preprocessing
Though data cleaning isn’t the most exciting facet of the modeling process it is arguably the most important. Data cleaning and preprocessing refer to the structuring and formatting of raw text data into something we can model. The care that goes into this process can make or break the statistical inferences of subsequent models. We will be using data from NLTK’s (a Python library for NLP) corpus module which consists of movie reviews along with their ground truth classifications (whether the review is positive or negative). Let’s get started by first importing the modules we’re going to need and downloading the necessary data…
