Topic Models

Topic models fall into the unsupervised learning bucket because, almost always, the topics being identified are not known in advance. So, no target exists on which we can perform regression or classification modeling. In terms of unsupervised learning, topic models most resemble clustering algorithms, specifically k-means clustering. You'll recall that, in k-means clustering, the number of clusters is established first, and then the model assigns each data point to one of the predetermined number of clusters. The same is generally true of topic models. We select the number of topics at the start, and then the model isolates the words that form that number of topics. This is a great jumping-off point for a high-level topic modeling overview.

Before that, let's check that the correct environment and libraries are installed and ready for use. The following table lists the required libraries and their main purposes:

Figure 7.2: Table showing...