You can play this game live by clicking here.
In the story, the player was about to visit a village haunted by werewolf. The four Twitter users the player input were other visitors in this village. But when darkness fell, one of them turned into a werewolf but the player didn't know who it was due to the chaos at night. The second morning, each of the Twitter users left a poem to describe what they saw last night. The player’s goal is to guess which poem was written by the werewolf by comparing the words and sentences they used.
The research question in this project is how to create data-driven poems based on a specific theme or within a storytelling context. Meanwhile, this project also imitates the traditional board game mechanism to explore the possibility of allowing people to interact with the computational literature even if it doesn’t make sense.
I designed the user interface and drew all of the illustrations to make the story more immersive and visually appealing to players.
As for how to create poems from Twitter data. Codebird.js helps the program to use Twitter API to retrieve data from Twitter accounts. The project mainly uses "statuses_userTimeline" API to fetch tweets from each user.
RiTa.js is a library used to manipulate the words and obtain their semantic information. First, this library will help the program to get noun words, verb words and adjective words from each user’s tweets. Then the project sets a context-free grammar for the generation of poems. To make sure the contents of the poems would be connected with the werewolf story, several templates were preset and left blanks for the twitter words to be filled in.
Here are two poem samples generated from different Twitter users: