MediaMind

MediaMind demonstrates that content-based filtering can be built into a fully functional, multi-domain discovery platform without requiring a large user base, proprietary training data, or any machine learning infrastructure. The weighted Jaccard similarity engine is intentionally simple — three features, three weights, one formula — but the results it produces are semantically meaningful. Items sharing high genre and tag overlap with a target consistently ranked at the top of the recommendation list in testing. Cross-domain results appeared in the right circumstances: when a book and an anime genuinely shared enough content features, the book surfaced. The 95% UAT completion rate and 4.2/5 satisfaction score are meaningful numbers because they came from evaluators with no prior exposure to the system. The single failure — one evaluator hesitating over the "Timepass" label — is the kind of small friction point you cannot anticipate in design but catch immediately in user testing, which is exactly why that testing phase exists. The work that would make the most difference in a future version is clear from the limitations: external API integration to automate catalogue growth, a behavioural feedback term in the similarity function, and server-side pagination. The architecture supports all three additions without requiring significant restructuring. Building them out is the next step.

[1]   P. Resnick, N. Iacovou, M. Suchak, P. Bergstrom, and J. Riedl,"GroupLens: An open architecture for collaborative filtering of netnews," in Proc. ACM Conf. on Computer-Supported Cooperative Work (CSCW), Chapel Hill, NC, 1994, pp. 175–186. [2]   J. S. Breese, D. Heckerman, and C. Kadie, "Empirical analysis ofpredictive algorithms for collaborative filtering," in Proc. 14th Conf. on Uncertainty in Artificial Intelligence (UAI), Madison, WI, 1998, pp. 43–52. [3]   M. J. Pazzani and D. Billsus, "Content-based recommendationsystems," in P. Brusilovsky, A. Kobsa, and W. Nejdl (Eds.), The Adaptive Web: Methods and Strategies of Web Personalization, LNCS vol. 4321, Springer, 2007, pp. 325–341. [4]   R. Burke, "Hybrid recommender systems: Survey and experiments,"User Modeling and User-Adapted Interaction, vol. 12, no. 4, pp. 331–370, Nov. 2002. [5]   Y. Koren, R. Bell, and C. Volinsky, "Matrix factorization techniquesfor recommender systems," IEEE Computer, vol. 42, no. 8, pp. 30–37, Aug. 2009. [6]   OpenJS Foundation, Node.js v18 LTS API Documentation. [Online]. Available: https://nodejs.org/docs/latest-v18.x/api/. [Accessed: Apr. 2025]. [7]   Oracle Corp., MySQL 8.0 Reference Manual. [Online]. Available: https://dev.mysql.com/doc/refman/8.0/en/. [Accessed: Apr. 2025]. [8]   Express.js Contributors, Express 5.x API Reference. [Online]. Available: https://expressjs.com/en/5x/api.html. [Accessed: Mar. 2025]. [9]   Nodemailer Contributors, Nodemailer — Node.js Mailer Documentation. [Online]. Available: https://nodemailer.com/about/. [Accessed: Apr. 2025]. [10] Auth0, "Introduction to JSON Web Tokens." [Online]. Available: https://jwt.io/introduction. [Accessed: Mar. 2025].