Summary

This chapter introduced many fundamental concepts, methods, and techniques for ML in the realm of text data. Then, we had the opportunity to apply this knowledge to solve a spam detection problem by incorporating two supervised ML algorithms. The content unfolded as a pipeline of different tasks, including text preprocessing, text representation, and classification. Comparing the performance of different models constitutes an integral part of this pipeline, and in the last part of the chapter, we dealt with explaining the relevant metrics. Hopefully, you should be able to apply the same process to any similar problem in the future.

Concluding the chapter, we need to make it clear that spam detection in modern deployments is not just a static binary classifier but resembles an adversarial situation. One party constantly tries to modify the messages to avoid detection, while the other party constantly tries to adapt its detection mechanisms to the new threat.

The next chapter expands on the ideas introduced in this chapter but focuses on more advanced techniques to perform topic classification.