Check out our project here.
Manga Recs is a project that recommends mangas (Japanese comic books) based on user-inputted similar mangas and keywords to query on. Users can log in and favorite certain mangas, which is an added UX bonus as users will have a personalized experience being able to arrange their favorite set of mangas. Furthermore, users can click on a certain manga title and see a detailed view of the manga, which includes the synopsis, as well as reviews given by actual users on MyAnimeList. The project makes use of the language retrieval, NLP, and other machine learning algorithms we learned in class.
The concepts learned in class that were used are: modified Jaccard similarity, basic text processing, vector space models, TF-IDF weighting, query expansion, word embeddings. The modified Jaccard similarity and TF-IDF weighting algorithms were used to measure similarity, and word embeddings were used to make sure that if similar keywords are inputted, similar results show up. We added a little bonus UI component where every matched query keyword is highlighed in yellow, and every matched similar keyword (through word embeddings) is highlighted in green.
Our web app's home page.
A query output. As shown above, word embeddings returns similar keywords to us.
The same query output, but scrolled down. As shown, matched words are colorcoded.
An example of what you would see if you clicked on a certain manga (i.e. it would show you the manga's synopsis and user reviews of it).
Our login page.