abhishyantkhare / book-summaries

Applying Encoder Decoder LSTMs to summarizing books. Most text summarization is focused on summarizing

Automatic Book Summaries

Most work in text summarization has revolved around summarizing short documents, such as articles or papers. However, books remain an important way that information is disseminated. Human made summaries of books have been incredibly important - ask anyone who has ever used Sparknotes to pass a Shakespeare reading quiz. Books in particular can benefit from summaries because they spend a lot of time diving into details, going over multiple examples, and spend time attempting to clarify a point. However, for anyone just wanting to get the main points from a book, it can be efficient to read a summary instead. This project attempts to create a neural network to automatically summarize books. I'm going to start with a LSTM encoder decoder structure, and improve upon it from there.

Data:

• The book text itself comes from Project Gutenberg, an incredibly valuable resource that provides free books: https://www.gutenberg.org
• The plot summaries come from the CMU corpus collected by David Bamman: http://www.cs.cmu.edu/~dbamman/booksummaries.html