The problem of collecting commonsense knowledge has been such a puzzle to me ever since I came into the lab. I think it's agreed by reserachers that personal commonsense would be a really useful kind of resource for recommendation system, personalized operating systems as well as applications, and so on. It's just that no one has ever come up with good ideas about collecting this kind of information, because, unlike general commonsense knowledge, only friends/family/related people around the target person can input the knowledge, so it requires a well-designed mechanism to efficiently collect this kind of data.
I was just chatting with one of my friends on line. My nickname was "guitar player+cook+scientist", and she thrown this message to me,"Then I'll be 'photographer+cook+fool-around'". I suddenly found that people tend to chat online about their lives using messenger systems. In other words, messenger systems are often used to exchange personal information between people who know each other (even though not all the time). Therefore, compared to websties or even blogs, where it's more like "Me and the general public", this kind of platform would probably be more suitable for collecting personal information.
Even so, I didn't see any possibiliy of finding a solution for collecting personal data, until the idea of
utilizing human computation' power via games came into my mind. I'm not sure if you'd agree with this, but around me, there're always bored people in my buddy list, trying to find others to play games with them. So if we can develop games - good games - on this platform, I bet it would be a fast way of collecting personal commonsense. The problem would then become: how we should design this game to do this job, such that, either symmetric or asymmetric, verification mechanism could be embedded to make sure the data's quality and authenticity.
The first idea came up to me is like this. When a user, A, wants to play a game, the system can find the other person, B, who shares a couple of friends with A, say C, D, and E. Then the game could ask various kinds of questions, not only about C, D, E to get their personal commonsense information, but also other topics to make the game more fun. Questions that are used to collect C, D, or E's personal commonsense could involve different types of media, such as their websites, online blogs, photo albums, etc, to make the game more interesting. For example, it could use one of E's photos and the mechanism of Peekaboo to gather the carried information.
I'm not talking about the specific look or usage of this game. I'm just wondering whether this kinda direction is worth trying. After all, I can't wait to use the personal knowledge base to make applications!!
延伸閱讀
