Send broadband support tickets to the right team from their words, with counts, TF-IDF and smoothing.
New to this? See the worked example firstNaive Bayes reads the words of a text and picks the most likely class. In these questions you send broadband support tickets to the Billing, Technical or Account team, turning text into word counts first.
Text classification comes up often in interviews; this set covers vocabularies, stop words, word pairs, smoothing and macro F1 for uneven classes.
If you are new to it, read the worked example first, then solve the questions in order: the early ones explore the data and train a first model, the middle ones test it on rows it has not seen, and the last ones tune it and turn the result into a business decision.