Two ways the Semantic Web may come to fruition are the top-down and bottom-up approaches. Using the bottom-up approach we would start from the bottom and work our way to the top by using a method like embedding RDF into Web documents to supply user agents with meta data. We are already seeing this type of action being taken by bloggers and other content creators. If we choose the top-down approach then we would start from the top and work our way down, using natural language processors to read existing Web documents and extract semantic metadata.

It's obvious which of the two approaches is gaining more traction. The top-down approach may require a technology that we are far from perfecting. Does this mean that bottom-up is the more appropriate course of action? Perhaps, but let's not discount the possibility of combining the two approaches. In a way we could use each approach for different sub-problems of the Semantic Web; Bottom-up to produce metadata for new Web documents and top-down for the billions of Web documents that will likely never be republished with metadata.

As it stands, bottom-up will likely lead the way for quite some time, but I expect to see the emergence of natural language processors in the next three to five years that will be able to tackle the enormous task of extracting semantic metadata from the Web's billions of documents. Are three to five years enough or am being a wishful thinker? Only three to five years will tell.