Take a sufficient sample of Gutenberg's digital book.

Create (random?!) samples of 200 partitions of the book.

Make sure each partition or record has 100 words.

Generalize the program so that you can replicate that for multiple books.

Maintain the label for each of the text segments or records or document, label them as a, b and c etc. as per the book they belong to.

Use Regular Expressions and Pandas to manipulate the data and serialize them.