Institutional Repository
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/0449081A-9A2D-4B88-8A8B-2C741FA40F67 | ||
| Year | 2020 | ||
| Type of Item | Diploma Work | ||
| License |
|
||
| Bibliographic Citation | Petros Portokalakis, "Providing personalized recommendations for interactive story generation", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2020 https://doi.org/10.26233/heallink.tuc.87378 | ||
| Appears in Collections |
Interactive narrative is a form of digital entertainment where players can create or influence a storyline through actions, usually by controlling the role of one (or more) characters in a virtual world. In story-based games or any other interactive story system in general, a drama manager is an omniscient agent that acts to guide the user through the story space. While drama managers tend to improve user enjoyment, they do not take into account the user’s preferences. In this thesis, we present a drama manager that tries to tackle the sequential recommendation problem, while taking into account user preferences. In order to create and maintain user engagement, we present a recommendation approach using probabilistic topic modeling, intertwined with reinforcement learning. We use the Latent Dirichlet Allocation topic modeling algorithm, applied in a Choose Your Own Adventure (CYOA) book to capture its latent topics. The key aspect of our drama manager is that we model the user herself as an evolving document represented by its respective mixture of latent topics; and which is appropriately updated every time the user consumes an item. Also, inspired by reinforcement learning literature, we introduce the use of variable learning rate for drama managers, directly associated with the user model updating, and based on the well-known “Winof Learn Fast” reinforcement learning method. The algorithm is trained using all the Wikipedia articles referring to books. We also provide an efficient parser for Wikipedia articles. Experimental evaluation results are promising, showing that our drama manager is capable of providing efficient recommendations to the user.
Copyright © DIAS 2013
