Infobright Blog

Hello,

I’m on a two-week holiday in Malaysia and Indonesia. It’s warm. Wonderfully warm compared to Poland and Canada. My five-year-old son is happy. So I’m happy too.

The question is how to compute my happiness from data? Imagine that I’m a piece of data and there’s quite a lot of information about me – I’m on holidays, ICE is getting popular, rough set book project proceeds, etc. So imagine a query like “find people who are happy” or “find average degree of happiness among people who are on holidays”.

I think there are at least two issues arising here. The first one is of course how to translate the available data definitions into the concept of happiness. Well, maybe happiness is not the best practical example but one can imagine some important features of multimedia or biomedical objects that are not directly expressible within the given data semantics.

The second issue is whether we really need (or want) to know exact quantity of “average degree of happiness”? Isn’t it enough to know some reasonable approximation? Actually, one can consider a similar question in more (so far) realistic applications. For example, I recently reviewed a very good paper wherein the Author discussed approximations of the results of OLAP queries. You are also surely aware of several approximate SQL projects. Honestly, I don’t think those projects were commercially successful. On the other hand, isn’t it just a matter of time when database users will have to accept approximate queries?

Motivation for approximations may be related to such aspects as: increasing complexity of concepts occurring in queries (like “happiness”), increasing complexity of data sources (where availability of data is limited), or, simply, increasing volumes of queried data. It’s actually possible to do some fancy approximate querying with ICE but I’ll keep it for one of the next posts (when I come back from holidays). In the meantime, as usual, I open a new thread in the forum. - What do you think about the future of approximate querying?

Best greetings,

Dominik