I’m an avid Youtube user because there’re a lot of good videos that always bring me something new. The only problem I have is its machine learning backed recommendation system which usually give me things totally out of sense, like videos in a language that I don’t understand, clickbaits and poorly made videos etc. This system is so sensitive that I may be bombasted with uninterested videos after I accidentally watch a video of a type.
I know Youtube has provided a way to tell I’m not interested in any video, but it affects so little comparing to how the system thinks you like one: I may have to repeatively tell Youtube 10+ times that I don’t like a certain type after clicking on one mistakenly. I understand it’s the stradgy that providing more to expand the possibility to keep me watching, but it’s simply getting annoying.
This is the idea of this project: creating a fair filtering system with machine learning to get rid of unrelated videos. The word fair here means that both positive and negative have the same effect on the system I’m building, unlike Youtube which always prefers likes than dislikes.
I’m really not a good programmer of writing anything big. It would be wiser to handle anything non-critical to some cloud platforms, so I only have to work on the system architecture and logics. I decide to go with AWS simply I have more experience with it. Azure and GCP are also good choices for completing same tasks.