… by Grigoris Antoniou and Frank van Harmelen

I finished reading it today. Ok, I admit, I finish flicking through it today. Simply not much to read.

The idea of Sematic Web comes from information retrieval instances like Googling. Current web pages/documents/videos are designed for human readability. If one wants to do a research that requires reading through a few pages, he may well google for it and summarize for himself. However, this approach does not scale.

Either the computer has to be smarter, or the documents have to be more machine-friendly. In the former approach, we use AI, and semantic web is for the latter.

I am not convinced by the points put forward in the book though. Imagine myself going through all the effort to explain my piece of information to the machine, I think the barrier to semantic web is too high, even though I acknowledge that any significant improvement on AI is nowhere near. With this in mind, the techniques in the book are of little interest to me.

However, if AI is not the solution, and semantic web is also not a solution, then how can we make search smarter? This question is of special interest, because we are what we search.

My take on this is to combine both. For datasets that are computer generated, we provide a form of standard format to represent data and metadata. In fact XML has already been in the way. The role of search engines are to adapt to each dataset in a smart way to “understand” the information they provide.

You might say that computer generated datasets are not enough. Well, if we can understand them alone that would have been a major step forward. But our engine built for this would not be isolated even if AI becomes superior so that they understand both computer generated and human generated information. Semantic web enhances the queries and the results. Google may have well tried that for their search engine.

Here is how it goes. You search for a set of keywords. Google found some ambiguity. For example, are you talking about Poseidon the God or Poseidon the game? Based on its ontology, it may ask you to confirm what you wanted. This would then give a more accurate search, since the terms you specified are now associated with the right occurrence profile. What more, is that the web pages information can now be pre-processed with AI, and their information is linked to some universal semantics. This way, the queries can be linked directly to the semantics obtained during preprocessing.

The above scenario is an instance of semantic web that can be executed in near future. I believe that search engines play an extremely important role in moulding the web. If your business stands out during searches, you are a step in front of your competitors. And to be able to stand out during searches, you have to be search-engine friendlier. If Google succeeds to plant the idea of semantic web into users, the semantic web standards will be able to jump through the barriers currently exist. But it’s only if Google tries and succeeds.