Applied Natural Language Processing in the Enterprise

Chapter 10. Visualization

In the previous chapter, we covered some of the most popular machine learning tools on the market today and shed light on the choices you’ll have to make for your ML software stack. Hopefully we didn’t leave you with a case of analysis paralysis.

But to ensure that you don’t get stuck in the “finding the right tool for the job” phase, let’s go a step further towards productionization and build a machine learning web app together. A web app is software that can be run from a web browser. This means your users don’t have to go through the extra step of installing your app before using it. More often than not, web apps also interact with web servers, which are remote computers that do more complicated things that you cannot expect a client to do, such as managing a database or, in our case, running inference on an NLP model.

Building a web app is useful because most humans do not derive satisfaction from scouring GitHub repos and dealing with CUDA out of memory errors. Creating a graphical user interface may seem like something that is not “real” NLP or deep learning. But having a simple graphical interface that’s accessible online is the most common way to have nontechnical users interact with your model. It’s also a great way to share your projects online, since most people (including deep learning researchers) would rather just click a link and see a demo than figure out what version of matplotlib to pip install because you forgot to include a requirements.txt ...