Inspiration
This game was inspired by old-school choose your own destiny games like Dungeons & Dragons or House of Danger, but with a small mobile game twist linked to it!
What it does
The game works exactly like those old school games, but entirely on one's phone! There will first be a prompt describing a certain situation, and the player will get to answer anything he or she or they want in return. Then, that answer will be analyzed by a NLP model in order to determine the most likely outcome of that decision. Finally, the game sends the outcome by text to let the player continue their story.
How we built it
The big two components of this project are Twilio and co:here. Using Twilio, the computer is able to first send a prompt to the player's phone, and the player is able to reply with their own unique answer. Once the computer receives that information back using Twilio, it is then uploaded to co:here in order to use their NLP model and classify the sentence. Depending on the key words used, the model will send back a code corresponding to the most likely outcome of that action. Finally, that "outcome" prompt is sent back to the player's phone by text to allow them to continue their adventure.
Challenges we ran into
The biggest challenge we ran into was setting up the two way communication between the computer and the phone. Indeed, the computer would upload a message to Twilio which would then send it to our phone. However, our phone had to upload its reply to Twilio, which then sent it to a public server. Finally, our computer had to access that server to collect the information. Needless to say, for a team of newbies, it was extremely difficult to learn the whole process of setting this system up, and we are very thankful to the mentors who were present to clarify the steps.
Accomplishments that we're proud of
We are very proud of setting up the two way communication system between our computer and our phone, using Twilio. Indeed, it was a hurdle to setup the public server from which our computer could collect the player's text reply. But as a team of amateurs, seeing our game play out so fluidly and by texting is really cool!
What we learned
We learned a lot on communications between the computer, the phone, and online servers like Twilio as well as the proxies in between. Setting up this whole system is far more familiar now. We also saw how adding varying kinds of training data affected the accuracy of the NLP model, as well as the accuracy of the classification.
What's next for Choose your Destiny - Mobile and NLP!
If time were not a factor, then we would implement an option at the very beginning to choose different stories to explore. We would also add images to the game's replies to make the adventure and story more vivid. Most importantly, with enough motivation, we would also work on and improve the NLP model with far more training data so that the classification is more accurate.
Log in or sign up for Devpost to join the conversation.