At first glance, text augmentation seems counterintuitive and problematic because the techniques inject errors into the text. Still, DL based on CNNs or RNNs recognizes patterns regardless of a few misspellings or synonym replacements. Furthermore, many published scholarly papers have described the benefits of text augmentation to increase prediction or forecast accuracy.
In Chapter 5, you learned about three Character augmentation techniques, OCR, Keyboard, and Random. In addition, the six Word augmentation techniques are the Misspell, Split, Random, Synonyms, Antonyms, and Reserved words. There are more text augmentation methods in the Nlgaug, NLTK, Gensim, TextBlob, and Augly libraries.
Implementing the text augmentation methods using a Python Notebook is deceptively simple. This is because Pluto built a solid foundation layer in Chapter 1 with an object-oriented class and learned how to extend the object with decorator as he discovered new augmentation techniques...