You are a new engineer at GTFlixTube deploying a Naive Bayes…
You are a new engineer at GTFlixTube deploying a Naive Bayes sentiment classifier for movie reviews! It works perfectly in testing, but in production, a user writes a review containing the exact phrase: “The acting was good, but the CGI was flabbergasting.” The word “flabbergasting” was never seen in the training data. Without smoothing, what will the model mathematically output as the raw probability for BOTH positive and negative sentiment?