Mysterio

Home Page
Game Page

Inspiration

Mysterio was inspired by NYT's Wordle, Google's Semantris, and Pokemon's "Who's that Pokemon".

What it does

When the player starts the game, they are shown the silhouette of a mystery image. The player's goal is to guess that image in as few tries as possible, and then subsequently solve as many mystery images as possible. The player can reveal a bit of the image, at a cost of some points. If the points for one round hit 0, the game is over.

How we built it

We used Google's Universal Sentence Encoder to convert the input and mystery words into vectors, as well as their synonyms and then calculated the angle in between the vectors to get a numerical, semantic difference between the two words. Then, the scaled difference will be displayed to the user.

Challenges we ran into

We struggled to get the vectorization step working since many models were not working for us. Many times we were able to get them to run, but oftentimes, the differences provided were inaccurate, and we had to rethink how we did the step. Eventually, we managed to use synonyms with the words, and a different vectorization model to gain a higher accuracy.

Accomplishments that we're proud of

We are very happy that we managed to complete the backend, especially the vectorization, and especially that we were able to integrate it with the frontend of the website.

What we learned

We learned how to utilize NLP models and vectors to gain a numerical, difference in meaning between the submitted word and the actual answer. We also learned how to integrate a Flask backend for a website with the overall backend of the program and a frontend site in HTML/CSS/JS.