Syntax and Themes: How Context Free Grammar Rules and Semantic Word Association Influence Book Success

Proceedings of Recent Advances in Natural Language Processing, pages 463–474 Sep 1–3, 2021. https://doi.org/10.26615/978-954-452-072-4_053 463 Syntax and Themes: How Context Free Grammar Rules and Semantic Word Association Influence Book Success Henry Gorelick1, Biddut Sarker Bijoy*2, Syeda Jannatus Saba*2, Sudipta Kar3, Md Saiful Islam2, 4, Mohammad Ruhul Amin1 1Fordham University 2Shahjalal University of Science and Technology 3Amazon Alexa AI 4University of Alberta hgorelick@fordham.edu, {biddut12,syeda06}@student.sust.edu, skar3@uh.edu, mdsaifu1@ualberta.ca, mamin17@fordham.edu Abstract In this paper, we attempt to improve upon the state-of-the-art in predicting a novel’s success by modeling the lexical semantic relationships of its contents. We created the largest dataset used in such a project containing lexical data from 17,962 books from Project Gutenberg. We utilized domain specific feature reduction techniques to implement the most accurate models to date for predicting book success, with our best model achieving an average accu- racy of 94.0%. By analyzing the model param- eters, we extracted the successful semantic re- lationships from books of 12 different genres. We finally mapped those semantic relations to a set of themes, as defined in Roget’s The- saurus and discovered the themes that success- ful books of a given genre prioritize. At the end of the paper, we further showed that our model demonstrate similar performance for book success prediction even when Goodreads rating was used instead of download count to measure success. 1 Introduction Since its publication in 1868, approximately 1.78 million copies of Louisa May Alcott’s Little Women have been sold, which equates to about 1,000 copies a month for 152 years. Every publisher in the industry hopes to find a manuscript that can sell even 10,000 copies in its lifetime. This begs the question: what makes Little Women a timeless suc- cess? Recently, researchers have attempted to use machine learning and natural language processing to answer this question, among others. Predicting the success of a novel by analyzing its content is a challenging research problem. Thou- sands of new books are published every year, and *Both authors contributed equally to this research. Figure 1: This figure represents the lexical production rules of context free grammar observed in both success- ful and unsuccessful books classified by Goodreads rat- ings. We present the count of 20 lexical rules, each normalized with respect to it’s highest occurrence in any book, for 10 successful and 10 unsuccessful books from CHILDREN genre. We see that certain lexical rules occur more frequently in successful books than unsuccessful books. only a fraction of them achieve wide popularity. Therefore, the ability to predict a book’s success prior to publication would be exceptionally useful to the publishing industry and enable editors to make better decisions. Many factors contribute to a book’s success including, but not limited to plot, setting, character development, etc. Additionally, there are some other factors that contribute to a book’s popularity that an author and publisher can- not control like the time when a book is published, the author’s reputation, and the marketing strategy. In this paper, we only focus on the content of the …

Content truncated to 500 words