Sunday, August 20, 2006

ESP Game

I just got a comment from Henk Poley on my posting Tagging Pictures is Bad Hunting where I said:
I am quite skeptical about finding possible solutions for this problem

Henk Poley:

Do you know the ESP Game? It's a game that solves this problem by letting people tag photos via a game interface.

Well, I saw the video with Luis von Ahn's brilliant lecture and I recommend all those interested in this problem to watch it.

I also used QTSaver to gather a few micro contents that will help us understand what it's about.
From Carnegie Mellon University comes this multiplayer Java applet simply called: The ESP Game. ESP Game is an experiment in human computation originally conceived by Luis von Ahn of Carnegie Mellon University.The idea behind the game is to use the computational power of humans to perform a task that computers cannot yet do (in this case, labeling images) by packaging the task as a game.

Humans are perfectly capable of it, but not necessarily willing. The ESP Game makes humans willing to perform this task by making a game out of it.Some users reportedly play over 40 hours a week. According to the game's creator, if the ESP Game gains popularity comparable to other online games, all images on Google Images could be labeled in just a few months.
The game throws up an image in a Java applet, then asks you and an anonymous "partner" elsewhere on the net to type in keywords until both of you have a word in common -- IOW, until you and a stranger can agree on a good label for the picture. Presumably, this is being added to a metadata database for the purpose of cataloguing all the images on the net.

"There's some meat to his idea," said his mentor, Manuel Blum, a CMU professor and a pioneer in the field of theoretical computer science. Producing word descriptions of images with the ESP Game is nice, of course, but the bigger idea is to entice people to cooperatively solve problems that defy electronic computers.
The ESP Game is a two-player game.You can’t communicate with your partner, and the only thing you have in common with them is that you can both see the same image.

The ESP Game: Labeling the Web - great model of using one type of interest (playing a game) to fulfill an unrelated need (creating an image database) -- it's the cornucopia of the uncommons!

Tagging Pictures is Bad Hunting

Searching pictures in any search engine is very clumsy and inaccurate. It depends totally on the description of the picture or on its tags (in case it is tagged on a photo sharing on line software like Flickr). The tags (or descriptions) are focused on the interests of the tagger, which don't always match the needs of the other WWW users.

E.g. I'm interested in Stars of David. There are dozens of pictures in Flickr which are tagged Menorah but include Stars of David. There are other items on these pictures which are not retrievable since the only thing that interested the tagger was the Menorah. Let's say that there are ten interesting items on each picture - imagine how much information is lost only because we (WWW users) don't tag properly.

This problem doesn't exist on text search. All the text "items" are retrievable.

I am quite skeptical about finding possible solutions for this problem, but I believe the more we use micro content anthologies the easier it will be to retrieve these "lost sheep". For example, I hope that in my micro content project people will find pictures that deal with Stars of David and were not tagged by the relevant tags in the place where I found them. The same goes for other websites that collect everything there is to know about a certain subject.

Thursday, August 03, 2006

New ideas to promote your Website

One of my readers asked my advice how to promote his Website. I told him I'm not an expert but he insisted. So here is what I wrote him based on my own experience:

1. Update - It is better to update 3 times a day with micro contents that once a week with macro contents.
2. Micro contents - If your website has already long articles you can break them down into many small units; publish each "chapter" in a separate page and erase the original (search engines punish you if you duplicate content).
3. comment on the discussion pages of Wikipedia and leave a link to your website - this will bring you many visitors since people turn first to Wikipedia when they need info and from there they go on their search (it helped me a lot - see the discussion page of the term Star of David).
4. Translate parts of your site to any other language on a separate blog. 
5. Traffic: read your traffic report and click the referrers' links to see who reads your articles. In case it's a blogger leave him a thank you comment with your link.
6. Letters to editors: send letters to editors of on line newspapers about the subjects you nurture in your website.