Learn to Write Popular Novels

For a long time, writing novels is traditionally regarded as personal and private activities. People record their unique experience and emotions and project it into complicated characters and elaborated stories. However, there are other kinds of novels, called popular novels or internet novels, that are written to satisfy readers’ emotional need and demand. Despite various world settings, many of such novels are simple in plots and character building. For example, the main character is mighty, powerful and morally right, and the story is how he/she defeats the evil characters and gain respect and affection from his colleage and lover(s). Although naive and low-quality in writing, such kind of story often receives much attention, since most of the readers are only set for casual reading. For the past few years, this demand has led to a substantially large market in China, where many websites (e.g. www.qidian.com) have been opened to attract both readers and writers. Writers, also called “internet writer”, are paid to write novels to satisfy the demand of readers. On the other hand, readers are paid to read novels, especially for those satisfying ones. To keep their audience and income, writers are constantly asked for intruging stories and massive daily updating, which is both timeconsuming and brain-twisting. As a result, only a very small number of writers can be successful. From a research perspective, one interesting question naturally arises. Since the plot is usually simple and repetitive, is that possible we can learn patterns from numerous novels published online and create our own? A trivial way would just steal others’ paragraphs and stitch them together, yet how to enforce entity and semantic consistency between paragraphs? In this paper, we try to learns a novel model from novel corpse on Internet using Markov Decision Process (MDP). This learned MDP enables us to find the patterns of authors and can be used to generate a new novel of our own. This approach can be applied to aforementioned websites to boost the quality of novels in the long run.