Dungeons and Dragons Fantasy Adventure Generation: Text Generation

We introduce the Forgotten Realms Wiki (FRW) data set and domain-specific natural language generation using FRW along with related analyses. Forgotten Realms is the de-facto default setting for the popular table-top role-playing game (RPG) Dungeons and Dragons (D&D). Wherein Forgotten Realms Wikia is a community-based knowledgebase dedicated to the Forgotten Realms setting. Domain-specific text generation is an emerging area in NLP in which the objective is to generate text that adheres to a specific domain, in the sense that the content generated should be semantically and pragmatically truthful to that domain. One of the reasons why domain-specific text generation is difficult compared to generic text generation is that in most cases this requires copious amounts of linguistic resources based on the domain in question. This hurdle is true even for fine-tuning a pre-trained model which relatively demands less amount of data than training a model from ground up. We perform D&D domain-specific natural language generation using the FRW corpus. Then we evaluate the semantics and pragmatics of the generated text in context to the domain by evaluating the named entity classification with respect to the lore of Forgotten Realms.