Inspiration
As people who frequently play Connections, the TAMU datathon prompt to make a bot solving them piqued our interest the most!
What it does
Picks groups of 4 words in NYT Connections style game.
How we built it
We used pre-trained LLMs, trained and validated them some more with our dataset, and matched words based on their vector similarities.
Challenges we ran into
We learned we were able to use online APIs halfway through, which changed the trajectory of our project.
Accomplishments that we're proud of
As freshmen, it's difficult to expect success or a working product from a hackathon, but our product works! At least with our data.
What we learned
Never be afraid to take on a challenge and always be flexible. If you're tired, sometimes the best way to progress is to take a break then continue on after.
What's next for Connections AI
Since Connections AI utilizes NLP through finding words with the most similarity, this process could be utilized in any scenario where word-matching is necessary. For instance in the medical field, matching drugs to diseases, or in education to help kids learn.
Log in or sign up for Devpost to join the conversation.