Google Code-in

Pick 4-5 novels that you may find interesting and try to build a classifier model that predicts the source of a given sentence. You will discover some interesting results if you change the name of characters or locations during inference (for example, replacing 'Tom Sawyer' by 'Harry Potter' in test set), or even make up a sentence on your own.

After that, try to train another classifier with all proper names masked (Hint: you can try using Named Entity Recognition to find the proper names). Compare and report your results. Can the second classifier capture the style of an author?