Nice, not sure how deep down the rabbit hole you've gone on this. However, if you do decide to go down the route of the tutorial, I'd recommend choosing a different data set as the one mentioned is movie reviews, essentially someone has labeled a ton on movie reviews which the model will use to classify against, the more relevant the type of content, the more relevant the sentiment usually. Fortunately, there are tons of similar labelings for various frameworks:
A method that is slightly less upfront work and could yield interesting result might be something like:
which would simply be an API request, you get 5,000 transactions free a month. I can't comment on the accuracy though as the last time I used it was a couple of years ago and was chat transcript classification.
While I'm here, I might as well ask for a chart too 👍. Could I get one for @quello as that account is much more active than my blog 🙈