Anime Recommendation & Keyword Cards

• 

  Anime Recommendations

• 

  Flipped Cards

• 

  Anime Selection

Inspiration

I combined anime recommendation concept with an educational concept (language learning).

What it does

You enter your favorite animes, and the app gives you personalized recommendations as interactive cards. When you flip a card, you'll see four words related to that anime—each displayed in English, Japanese, and Romaji. The words are generated by Gemini, blending fun and language learning.

How we built it

I've already collected user data and started planning a project around it. I came across a BERT-based recommendation system that was originally trained on the MovieLens dataset. So, I adapted my dataset to the same format, trained the model, and built an anime recommendation system.

For anime information, I used data from the Anime Offline Database: https://github.com/manami-project/anime-offline-database

For alternative anime titles, I used this MyAnimeList dataset on Kaggle: https://www.kaggle.com/datasets/trushankvashikar/myanimelist-anime-dataset-till-june-2025

Challenges we ran into

None

Accomplishments that we're proud of

There aren’t many online anime recommender apps that use BERT, despite its impressive performance—so I’m proud to have brought one to life and made it available online.

website shutted off

Built With

• gemini
• python
• pytorch
• railway