
Flesch Kincaid Readability Tests
Data Skeptic
00:00
Natural Language Processing
The main area that i find this interesting for is as a tool in the feature engineering process. So when you get some text data, you can't directly feed that into an algarithm because it's just text. Most machine learning approaches need fixed length of input. This embeddings solution will take an arbitrary amount of text and compute a fixed length vector - something like bert or g p three. Those resulting vectors are spookily useful for machine learning augarithms. They find useful features in those large dimensional imbettings, but they might also find useful features in things like the flesh readability tests.
Transcript
Play full episode