Watchlist Recommendation Engine
Today I get to tell you all about a killer new feature to our online Watchlist. This is a feature I literally began writing more than 3 years ago. Back then I began writing a relationship function for our TV Show watchlists that allowed our software to group shows together based on how commonly they appeared together in users Watchlists so that if you had some of the shows in this group but not all of them the recommendation system could recommend you get the other shows it believes should be in that group.
When I started writing this system all those years ago it became quickly apparant how difficult it would be. First of all you need a very large piece of data to make the relationships between shows more than just one offs. You also need to understand that some shows shouldn’t be grouped together, just because someone likes The Simpsons and Game of Thrones doesn’t mean other people who like dramas like Game of Thrones would enjoy Cartoons like The Simpsons. If you have enough data it is likely but it’s still just a high probability guess.
It was for this reason that several years ago I began collecting data. I turned on an extremely verbose show relationship engine and every time someone searched for a show, looked up a show using the !show command or added a show to their watchlist that data was saved with their ID, correlated and ranked. The more effort the user spent to seek that show (for example adding it to their watchlist) the higher the rank that show gained within all that users show queries and thus higher its relationship status.
Now with 3 years of data piling up I was going through my database and noticed this mountain of data and I realised it is probably about time I did something with it. So a few days ago I began writing the process that will make all this data meaningful and accessible and then I completed that about 8 hours ago and since then to now I’ve been coding the interfaces for the website and the function for the watchlist system to access the recommendations.
Now when you open your Watchlist you will see recommendations at the top in green. Right now you cannot turn this off but I will be adding a toggle to hide it if you don’t care for it. You will see two recommendations at any one time and below the recommendations you have buttons to add the show to your Watchlist, view information about that show on Wikipedia and watch something related to the show from YouTube. The YouTube button relies on YouTubes searching system so it may not always give a trailer for the show and instead a review or a fan-video but it should at-least give you a good sense of what the show is about.
So yeah, recommendations. I’m super super happy to finally be able to deliver this feature. It’s excellent and this is also the first Watchlist feature that is exclusive to the web interface. I made the decision earlier in the week not to add this to the in-room Watchlist interface and hopefully it will drive even higher adoption of the website which I feel is the best way to experience your Watchlist.
Thanks for reading!