Movie/Tv Show recommendation

Inspiration

Our tool recommends new TV shows and movies based on your preferences regarding previous media that you have consumed.

What it does

Using text embeddings, key features about the show are mapped to a 1024 vector space. Potential shows are ranked by considering the distance to favored points and away from disliked shows. We used a Euclidean distance metric, but this metric can easily be modified.

How we built it

Flask + HTML,CSS, Javascript for front end. For the word embeddings, Ollama's mxbai API was used. Pandas and CSV were used to store these embeddings and for efficient access.

Challenges we ran into

Large file storage with CSV, and time optimization of calculating vector distances

Accomplishments that we're proud of

The embeddings and cost function were able to successfully generate recommendations similar to the ones the user preferred while staying away from shows that users disliked.

What we learned

Back end development with Flask, Event handling, UI/UX design, word embeddings, and sentiment analysis.

What's next for Movie/Tv Show recommendation

Sort by genre, randomly select a movie or tv show and give information about it. Allowing specific user feedback about a show to influence recommendations via a Transformer.

Built With

• css
• csv
• flask
• html
• javascript
• ollama
• pandas
• python