As it stands today, there are two primary approaches to categorizing content for the Web 3.0 objectives.

The first is the "bottom up" approach, which involves embedding semantical annotations (meta-data) right into the data. Primary technologies involved in the re-annotation of existing web content are based on RDF organizational schemas.  New ontologies are emergenging to simplify the process and include languages such as Turtle. Many research driven pursuits in the semantic web are focused on implementing these technologies into the Internet today in order to ease the organization, navigation, and structure of the data contained on the web today.

The second approach is "top down" and relies on analyzing existing information across the Internet and considers using natural language processing to understand and interpret information.  Many commercial interests have developed from trying to "understand" the content of the web.  Various mechanisms including tagging and extensive databasing are being used.  

However, what will likely emerge, and what is currently emerging, are more hybrid models that integrate both the bottom up approach where new information, including Web 2.0 content of the future, begins to take the underlying structure of the technologies we have discussed, which existing data across the Internet, and that data that remains resistent to these new technologies will rely on more sophisticated NLP programming for compiling and data mining.