Instruction:
Dr. Michal Halamish

This project is a tool for assisting the Dungeon Master in a Dungeons & DragonsTM tabletop role-playing game. Dungeons & DragonsTM is a storytelling game that takes place in an imaginary world. The Dungeon Master (DM) and the players play the game together by describing the events and actions using natural language. The DM’s role is to build the imaginary world, set the scene for the players (who each play a character), respond to their in-game actions and advance the story by creatively guiding the in-play dynamic.

D&D players willing to take on the DM role and run a game are often hard to find because Dungeon Mastering is a complex craft that requires creativity and quick thinking. The Dungeon Master often has to analyze situations and make game-changing decisions on the spot, while the players only have to play their own characters. This project aims to make the DM’s job easier by helping them make decisions that advance the game and analyzing the current in-game situation.

Natural Language Processing and creativity are highly demanding tasks for a computer. After all, computer algorithms have no understanding of the words’ meaning and pragmatic context. To help Dungeon Masters, the students used advanced algorithms capable of automated deduction, trained on a large set of examples.

They successfully implemented the algorithm and developed the tool by responding to feedback from very experienced Dungeon Masters. As a result, the system was well-received by its main target audience.

The project consists of three parts:

• A machine learning algorithm that analyzes text input from the Dungeon Master and catalogs it to choose the next step. A tool for us and other developers.
• An online platform for running text-based D&D campaigns, where the Dungeon Master can use our system discreetly.
• A game server that connects the machine learning model to the online platform and displays the information produced by the model in a convenient format.