Almost a year ago now, I wrote a post here on "Human Assisted Computing: Putting CAPTCHAs to Work". Among other things, the post referred to a game created by Luis van Ahn called "The ESP Game". In the game, two players (sat at oppostie ends of the web!) had to label the same image in a consistent way, the idea being that they could help train search engines to find images on the web more effectively (for more information, see the article "Games With a Purpose").
A week or two ago, the Games with a Purpose website launched, containing four new games alongside The ESP Game. As with the ESP game, all the games are two player games, that pair people who are unknown to each other across the web.
- Tag a Tune: both players hear a tune and have to describe it to each other; they then have to decide whether they are listening to the same tune. The long term reason - help a search engine learn more ways of finding songs (for example, whether they are happy or sad).
- Verbosity: players take it in turns to describe a secret word to each other - one person describes the word, the other has to guess it.
- Squigl: both players seem the same image, and are presented with a word; they each trace round the object described by the word as it appears in the picture.
- Matchin: two players are shown the same image; each player picks the image they like the most.
In fact, this last game - Matchin - reminds me of an image based 'personality profiling' service - "Visual DNA". In this 'game' you are presented with a series of image collections, and have to select one that best answers a question or statement you are also shown. After completing the exercise, you are presented with a 'personality profile' based on your selections. The idea of this service is - people matching (that is, it's a dating website...)
Intrigued by the idea of using people power to help search engines learn, I've started thinking about building my own "game with a purpose". And this is what I've come up with: Trendspotting.
Trendspotting uses the Google Trends web service to encourage people to identify search query terms that in some way relate to each other, or that have some sort of 'periodicity' - that is, that are more popular at particular times of the year.
For example, some search terms follow natural phenomena:
Google Trends: 'full moon'
Some follow 'events':
Google Trends: 'tennis'
The idea I'm working towards with the game is that players can submit trends that show a very definites periodicity - or repeatability - year on year, or maybe show how certain terms 'lead' or 'lag' one another in certain well defined ways. (I did think 'flowers' might be popular a day or two before "valentine" for example...!)