Title:Tropes in films: an initial analysis

Abstract:TVTropes is a wiki that describes tropes and which ones are used in which artistic work. We are mostly interested in films, so after releasing the TropeScraper Python module that extracts data from this site, in this report we use scraped information to describe statistically how tropes and films are related to each other and how these relations evolve in time. In order to do so, we generated a dataset through the tool TropeScraper in April 2020. We have compared it to the latest snapshot of DB Tropes, a dataset covering the same site and published in July 2016, providing descriptive analysis, studying the fundamental differences and addressing the evolution of the wiki in terms of the number of tropes, the number of films and connections. The results show that the number of tropes and films doubled their value and quadrupled their relations, and films are, at large, better described in terms of tropes. However, while the types of films with the most tropes has not changed significantly in years, the list of most popular tropes has. This outcome can help on shedding some light on how popular tropes evolve, which ones become more popular or fade away, and in general how a set of tropes represents a film and might be a key to its success. The dataset generated, the information extracted, and the summaries provided are useful resources for any research involving films and tropes. They can provide proper context and explanations about the behaviour of models built on top of the dataset, including the generation of new content or its use in machine learning.