Recently in Thinking Category

I found a cool search engine - SearchMe (a gold supporter of the conference) on the website of CIKM 2008. It uses a totally different way to present the search results. Compared with snippets, picture slides is much more intuitionistic and friendly to users. Users can easily make an initial judgement whether the contents of a web page can satisfy their requirements.
To be honest, the most interesting feature attracts me is the classification of the returned web pages. When we input the keywords, a classification is instantly shown on the top of the slide. This feature is very useful for effectively filtering out the irrelevant results. Of course, the accuracy of classification should be improved (I think it mainly depends on the research of web mining), but this is a good direction.

For sure, Dapper can help our web sites convey semantic information or meta data and make our web sites more structured. But I doubt whether it can help the search engine improve the accuracy or precision of the results of a keyword search. Have a look at the keyword search over relational database or XML database. The data is structured, but the results of a keyword search is not ideal. Therefore, the precision of results is not determined by whether a web site is semantified. Semantifying is just the first step.
Users express their intention through several keywords, but these keywords convey too limited semantics, so it is very difficult to find out what information is really wanted by the users. If this problem can not be properly solved, the semantic information conveyed by the web sites may be not very helpful. Therefore, when our web sites become semantified, we also need a new simple query pattern to make the keywords more semantic.