Turtle Soup Generator

Situation puzzle generation workflow

Situation puzzle is a popular text-based interactive game. In such a game, there is a moderator and a player, or a group of players. The moderator has a complete story in mind and provides the player(s) with the beginning and the ending of a story; the player(s), on the other hand, are expected to figure out the step-by-step details of story development by asking the moderator ”yes-or-no” questions about the story. The game would end either when the player(s) figure out the complete story line or when a preset number of ques- tions are asked before complete the story. The challenging and intriguing part of this game is that the player(s) need to come up with a story with only the beginning and the ending while also make sure the logical relations in the story align with commonsense.

We propose to build a system (TurtleSoup) that not only automatically generate high-quality story for a situation puz- zle game but also handles the job of the moderator to in- teract with player(s). With minimum human intervention, TurtleSoup is capable of providing comprehensive interac- tive experience that can be entertaining.

Generation and Question Answering demo
Generation and Question Answering demo
Xinyue Wang
Xinyue Wang
Graduate Student of Bioengineering

My research interests include Causal Discovery, Causal Representation Learning and Machine Learning Application