Jinni - Movie Recommendations Site

Music recommendation services like Last.fm do a fairly good job of picking music based on your listening habits, and comparing those sets with similar users. Frustrated with trying to figure out what to rent from the local video store, I looked into similar service for movies. I started out back at IMDB.com, which is ok if you really only want a homogenized general rating of a film. Basically, a high rating will ensure a movie's general public appeal, but not necessarily whether I would personally like it. Like musical tastes, our tastes in movies is a pretty subjective and personal thing. Last.fm's approach of just mindlessly processing consumption habits (number of times listened to), isn't likely to work for movies. A movie lasts a fair bit longer than a song, so I'm unlikely to watch 2001: A Space Oddessy over and over enough to automatically rank it as a favourite film. Similarly, I might have seen Anchor Man based on a friend's recommendation and really disliked it, which would mean I've seen it almost as many times as Dancer in the Dark, one of my favourite films (if only because its content is so heartwrenching to watch). Enter Jinni.com, the online movie recommendation service. I've only been playing with it a couple of days, but there's a lot to like (and a quite a bit to improve yet). It appears that Jinni uses a pretty rich ontological system (contextual relationships between different terms to descibe characteristics of the media in question) to categorize films. You start out by ranking a bunch of films that they've selected to create profiles of different 'taste types'... Social Animal, The Individualist, etc. This information is then used to create a set of recommendations for you. The really cool power of Jinni though is its contextual search, where you can type in words like "historical drama gripping bleak" that describe different characteristics of the kind of movie you want to see and it gives you recommendations based on how movies in the database have been tagged.

It's all pretty fun to play with and the interface is very Web 2.0 (more on that in a  bit). One issue I had, was that the recommendations based on my 'Taste Types' and the tag-based 'ontological' results seem to exist in separate domains. The results of the tag-based search don't seem to be weighted by my 'taste types' or take into account movies that I've already seen and rated. Perhaps there could be an option to use personal taste weightings, or turn them off. My main qualms with the site were with its interface. The user interface makes use of a lot of rich-media tricks like dynamic mouse-overs and interactive button elements. This, however, seems to put a fairly high load on the Jinni server as well as my bandwidth. The responsiveness when making selections can be very sluggish as a result, and I had a couple browser freezes while using the site. A more specific niggle is that movie 'synopsis' text in the mouse-over is about 1.5 short lines of text, which often isn't enough to figure out what the movie is about. Clicking on 'more' (eventually) takes you to the movie's entry page... where you actually have to click 'more' again(!) to expand the synopsis.

From what I've seen though, Jinni is real step forward in figuring out what movie I want to watch. It's still in beta, so I trust that many of the technical issues can be worked out, and additional usability features brought online. If you'd like to try it out yourself, fire me off an email and I'll set you up with a beta invite.

Posted on January 30, 2009 and filed under Flicks.