Predict NYT bestsellers with wordpiece tokenization 

Hello Julia, I'm glad I've found your channel. Your book on text mining helped me immensely with the realization of my bachelor thesis and was a great introduction to new methods for someone like me, who studied in a field, where the use of digital tools always is one or two decades behind. This way of gaining insight on things is a lot of fun. I'm almost sad, that I didn't know about this earlier. Thank you, keep it up and have a great day.

Missed you Julia you really do great stuff and always have insightful content.

Thanks Julia, long time no see, pls keep up good work, I am still learning tidymodels through out your screencast everyday and hopefully can manage it at some point. cheers

Thanks Julia! Would be interested if you can improve prediction accuracy using a deep learning model via tidymodels

I was a little confused about word pieces, so here's the explanation from the package documentation: 1) Put spaces around punctuation. 2) For each resulting word, if the word is found in the WordPiece vocabulary, keep it as-is. If not, starting from the beginning, pull off the biggest piece that is in the vocabulary, and prefix "##" to the remaining piece. Repeat until the entire word is represented by pieces from the vocabulary, if possible. 3) If the word can't be represented by vocabulary pieces, or if it exceeds a certain length, replace it with a specified "unknown" token. It seems as though this will be effective for finding name suffixes like "-son": if "John" is in the vocabulary but "Johnson" is not, then the latter will be tokenized as "John" and "##son". I suppose this won't go so well with prefixes like "O'-", "Mac", "de", "van", etc.

You are awesome!! Thanks!!

Hello Julia, is it necessary to set the seed more than once for reproducibility? If I set it only once won't it be reused?

If anyone is looking for a new pen name, you might consider Janet F. Steel