Anime Recommendation System

Inspiration

We're really passionate about anime and wanted to create a tool to help introduce people to new anime.

How We Built It

We identified a list of publically accessible information about anime that could be used to help inform a user about what to watch next, including genres, tags, and descriptions. Using FastText and this Kaggle dataset, we were able to create a recommendation system that weighted shows a user has previously watched based on the score the user gave it. These weighted shows were used to identify other shows in the dataset that the user could then watch next. The recommender itself is a Python script that generates an ordered list of recommendations that is intuitive to read.

What We Learned

We learned about multiple natural language processing techniques while reviewing how to use the descriptions as a data point of comparison between anime. Additionally, we learned a lot more about how to vectorize elements with multiple different parameters more efficiently.

What's Next for Anime Recommendation System

If we had more time, we would have liked to add a webpage frontend that would make this project more easily accessible, as well as utilize a faster model to output recommendations more quickly. Futhermore, there is room for growth by using datasets than are not drawn purely from Anilist.

Built With

• anilist
• kaggle
• python