Abstract: Book adaptation is an important innovation source for movie production. Similar to other evolutionary product innovations (e.g., smartphones vs. dumbphones), adapting a book into a movie provides an opportunity to reach a larger audience, but at the risk that the adaptation may not be accepted and get negative word-of-mouth (WOM). This raises important questions from the perspectives of the movie producer/manager: What books are suitable to be turned into movies? How much should one “innovate” when bringing the book to screen? Do consumers focus on similar topics while watching the movie vis-a-vis reading the book? How such similarity is related to the success of adaptation? In this project, we intend to answer these questions by leveraging machine learning tools, especially natural language processing methods such as Latent Dirichlet Allocation (LDA) and Doc2Vec to gain insights into what topics readers/moviegoers focus on when they read the book or watch its movie version. Also, we will numerically represent how similar the movies are compared to their original books, both objectively (based on movie scripts and book content) and subjectively (based on consumers reviews). Such quantitative representations will be then taken into further statistical analyses to answer the questions mentioned above.