Semantic Web in the Startup School

Startup School was excellent! I met Paul Graham and Peter Norvig personally, the two people I was mostly looking forward to meet and ask them questions. What kind of questions? Well, you guessed it. About the Semantic Web.

Paul Graham said: "Semantic Web is a red flag". I was shocked! "Explain what you are doing. Don't say it is the Semantic Web. Once your application is ready, then say its part of the Semantic Web", said Graham. And he is totally right. People have gone on saying what is the Semantic Web, and haven't been able to explain it very well. The best way to explain anything is with an example, so we are still waiting for the killer app. Kingsley Idehen, from OpenLink Software, had a very good response. The lesson: start building Semantic Web apps fast. Start now! I already started!

The climax of the day was Peter Norvig. I summarize his talk in the following: use existing data to do better. He talked about improving image search with existing data on the web and continued by explaining a text segmentation problem. One example, where existing data could be useful, is to create a set of words, given a couple of words as input. For example, given lion, tiger, bear, you could search the web and get related words like elephant, monkey (other animals), but the words wood, musical, toddler would appear. Why? In my words, its the syntactic problem of the Web. If you have a text about animals in a musical wood box for toddlers, well obviously these words may have some "syntactic" relationship with animals. We can assume that there is a relationship, but actually, semantically, there is no relationship.

Please check his slides and videos, to view his talk. But the best thing, is that I was able to make a question. To hear my question and his answer, look for 20:55 or -5:01. He said, "there are already a lot of semantics in the web", and then points to the examples of his talk. I completely disagree! That is an example of the syntactic web, and not the semantic web. You can extract semantics out of the syntax, but that doesn't mean that it is accurate. Peter Norvig is known for being against the Semantic Web, well, that is what most people consider. This post here and here will explain.

But to my surprise, he actually did give Semantic Web the credit it deserves. He considered "great that communities can get together and agree ..like in biogenomics area. That is a good model. Its a community that has a need." He considers that academic and friendly instead of competitive and "those kinds of attributes, I think that will coalesce together and grow. But for the world as a whole, you wont necessary see that." Aha! So it will work for the bio-community (as it is now) but for the WWW, it won't?

He was about to finish answering my question, but I was able to ask him about the Linked Data. He answered saying that "we need more flexible ways to link data." And I agree with him completely! But that is what the research community is working on. Personally, I work on integrating the data in relational databases (that are the biggest source of data on the WWW) with the Semantic Web automatically. Norvig continued saying that the "Semantic Web is a treaty between parties to say we are going to agree on this particular ontology. The more treaties, the better. You shouldnt expect everbody to lay down their arms on this." In my opinion the linked data image is a demonstration that more and more people are getting together.

So is Google interested in the Semantic Web? "We are interested", said Norvig. He continued, "we do want to support it and see where it goes, so far, we don't see it making an impact."

After his presentation, I was able to talk to him personally. I told him that before he seemed that he was against it, but now, he is even kind of for it. He said. "eehhh". I will take that as a "we are secretly doing work on the Semantic Web, but not publicizing anything until we have good results."

The cool thing was that everybody in the auditorium now new that I was a Semantic Web guy, so immediately I was able to meet the people I was looking forward to meet: people interested in Semantic Web. Unfortunately it was a small number, but at least it wasn't zero. I was glad to finally meet somebody from TopQuadrant, which was Keefe. I also met Waleed Abdulla from Zooov, which was one of the persons who initially started to work on SPARQL four years ago. And it was great to meet people interested in the Semantic Web like Daniel from Biographicon and Sebastian from Intalio, and how it can help their ideas and business.