Connections AI

Inspiration

We were inspired by large language models to create a connections ai.

What it does

It attempts to solve games of connections by returning guessing as lists of 4 words. It takes in a list of 16 characters and based on how successful previous guesses were it attempts to solve future guesses

How we built it

We used a large language model, trained it to predict words based on category, context as well as existing datasets.

Challenges we ran into

We had issues with the prediction of the words as it consistently confused context.

Accomplishments that we're proud of

We're proud of achieving 90/90 on the autograder based on set and our fallback LLM is also really cool.

What we learned

We learned a lot about training a model from set data and how to contextualize synonyms

What's next for Connections AI

We plan on making it more accurate in the future!

Built With

• chatgpt
• pandas
• sentence-recognition
• spacy