While pondering over the 'meanings' of Semantic Web, BBC mentions a semantic search engine called Swoogle (On Wikipedia) developed by the UMBC. I registered with it but couldn't search as probably the site is overloaded. Kept giving me an error. The picture I got is that Swoogle crawls on web documents and categorizes information by 'reasoning' about it. More technically, it builds ontologies. Simply put it identifies and classifies objects that it can find from a web document and tries building relationships between them.
Example: When it scans a page on gadgets, it first identifies that it is a page about gadgets and starts classifying the gadgets that are listed. Suppose the gadgets include a cellphone and a PDA, it documents the relation between them, probably more, like how they can communicate to each other. So on and so forth.
That's as far I can understand. 🙂
I remembered Kosmix which was launched a while ago. Was started by some Indian brains if I am right. They have now expanded to video games & finance. (Their initial domain specific searches were on US politics, Health, Travel)
Previewseek had taken their site off after some preview last year. My friend had a post then when we tried a few searches on it.
Both Kosmix and Previewseek seem similar in their results. Previewseek was more general and hence very fickle in its classification of results. A search on "Sprain" in Kosmix Health would yield among other things.. the basic info, causes, symptoms, treatments etc. A search under politics would yield items classified under liberal, conservative views etc. Definitely a way forward, but I believe most of it was data restructuring. I donno if its a long term solution because a) You cannot do that every domain. Only very important domains merit such hard effort. (I am assuming still that it was data restructuring along with some amount of algorithmic effort) b) Domains get created every day and information changes constantly.