Connections AI

Inspiration

While playing the New York Times Connections game, our team faced difficulties solving the puzzle and we realized that our vocabulary bank was too small. As a team of Computer Science students, we decided to leverage our technical skills to build a machine-learning model to solve this puzzle for us.

What it does

Our machine-learning model takes a set of words as the input and finds themes between words. The process of doing so is as follows

1. Preprocess New York Times word dataset
2. Use spaCy to tokenize the words for semantic analysis
3. Train our Word2Vec model
4. Perform hyperparameter tuning for our model

How we built it

• Caffeine
• Sheer determination
• scikit-learn
• spaCy
• Gensim

Challenges we ran into

• Struggled to find sufficient training data
• everything else was also a struggle

Accomplishments that we're proud of

• Given the difficulty of this challenge, we are proud that we stepped out of our comfort zones to try this challenge.

What we learned

• Reading research papers
• Using different Python libraries to build our model

What's next for Connections AI

• Destroy this game with our ML model

Built With

• gensim
• matplotlib
• python
• scikit-learn
• spacy