I’ve had a grand plan for a direct message correction system in the works for months now but I dug myself into a rather deep hole. Taking a step back I realised I was massively overcomplicating things and decided to start again. The most common problem people have is with the categorisation going awry so that’s where I decided to start… So, I give you: direct message corrections! It’s still a bit too simple for my liking but it’s safe to play with. Here’s what you need to do:
d oo5 hollywoodland is a movie
d oo5 third by portishead is an album
d oo5 Chris Helme was a gig
Dead simple… The corrections are grabbed every five minutes (ish) so don’t worry if nothing happens immediately.
Comments (View)
Twitter isn’t letting @oo5 grab reviews at the moment… New ones will turn up as soon as they get the service fixed.
Comments (View)
Search has exposed a little Easter egg that I’ve been sitting on for some time. A while back I moved contributors’ locations (according to Twitter) onto the reviews. I figured that before long people would be using services connected to Fireeagle or apps like Brightkite and Twinkle to keep their Twitter location up to date… So if you use any of those services it means that @oo5 has been automatically geo-tagging your reviews.
I’ve not had time to do anything with that data yet but if you want to have a play, pop some cities into the search box and see what comes back.
Comments (View)
New this week: Speed Racer. Jealous of: anyone who was at the ATP festival over the weekend.
Comments (View)
Film of the week: Iron Man with a 4.1 average. Album of the week: White Goddess By James Blackshaw with a 5.0 average.
Comments (View)
At the moment @oo5 is a black hole. You review stuff, it gets sucked into oo5.whatiminto.com and then… nothing. I’ve been pondering this problem since I started the project.
For me, reviews are more permanent snippets of communication than standard tweets. They’re also more structured. I wanted to be able to store them somewhere - and in a way - that would make them easy to get at later. So my first priority was to appeal to people like me: people who wanted a permanent record of what gigs they’ve seen, movies they’ve watched, places they’ve eaten. Out of Five could be a micro-micro-blog derived from Twitter, and promoting Contributors to the main navigation was intended to promote this.
I also wanted to make that data as easy to extract and reuse as I could. People with a certain degree of technical expertise might want to pull their reviews into their own websites or applications. To those ends, I made sure that RSS and XML feeds were available for all the useful URLs to anyone familiar with RESTful resources. I also marked up the reviews as microformats to give anyone with the will another axis of access.
What I haven’t done yet is establish any real value for the rest of the world. If you’re not a compulsive obsessive diarist, a programmer or a blogger then what’s in it for you?
Adding some form of recommendations seems logical.
The most obvious way of doing recommendations is by using your social network. I can get your friends from Twitter and tell you what they’ve been reviewing… But it’s Twitter and you’ve probably seen their tweets already. Result? Not very compelling.
The scariest is working out a mathematical recommender. “People who liked the same stuff as you also liked…” The catch is that I’m not really a programmer and maths has never been my strong suit. That said, a quick flick through the Programming Collective Intelligence book has turned up a vector space algorithm that I could port. Interesting.
The simplest would be to send out a chart every week. If you mostly review films I could post you the week’s top films that you’ve not reviewed, for example. It’s simple, which is good but I wonder how much fun it’d actually be for people. Would it make it worth your while? I’m not sure.
At the moment I’m coming down on the side of simplicity but I would absolutely love to be able to recommend someone something that they found genuinely interesting. Hmmmmmm.
Comments (View)
