Oxford Internet Institute talk by Tim Berners-Lee
These notes attempt to capture some key points that I drew from Tim Berners-Lee's OII presentation on 14 Mar 2006 . (For me, this talk was probably "preaching to the choir", as I have been involved with Semantic Web standardization activities for several years, but this wasn't simply a rerun of material that I've heard all before...)
The talk was largely focused on building a "web of data" that could be "browsed" by computers as well as by people.
These notes were initially made from memory on the morning following the talk. I have since cross-referenced them to the slides of the talk, but have not otherwise added much significant content. As such, these notes reflect what (for me) stood out as major points of the talk.
1. Semantic Web vision
I felt that the vision of the Semantic Web has greatly matured over the past 2-3 years. Due in large part, I believe, to the near standardization of SPARQL (the Semantic Web query language), there is greater focus on integrating existing data sources through the RDF data model rather than combining information expressed using RDF syntax. This is, I think, a far more credible view than some of the earlier approaches. In introducing RDF, greater emphasis was placed on its relationship with the relational data model, rather than the purely graph-based approach that has previously been offered. The graph overlay of combined tabular (relational) and tree (XML) data was illustrative of this idea.
Also emphasized is easy merging of RDF data:
2. The RDF bus
(This bit was added on reviewing the slides as I think it's a very important principle. However, it wasn't something I remembered when writing down my original notes.)
Semantic Web applications were presented as "clients of the RDF bus". Existing applications can be part of the Semantic Web by providing adapters so that their data can be presented in RDF form, without necessarily converting the application itself. (For me, this echos a crucial observation made several years ago by Brian McBride of HP Labs, that we should be asking not how we can convert applications to use RDF, but rather how we can bring the benefits of RDF data processing to existing applications.)
3. Technical rules and social rules
The interplay of technical and social rules is a crucial aspect of the web, often overlooked by technical developers (I've been there!). But it is a two-way interaction. I was particularly taken by the observations that incremental changes to the technical rules could bring about large (and unpredictable) changes in social behaviour. That seems obvious in hindsight, except maybe how small some of the incremental technical changes have been in relation to their social effect.
4. Ontology economics
There was an excellent slide about a proposed relationship between the amortized cost of developing an ontology related to the number of people using (i.e. agreeing to use) that ontology. It has been a long-running criticism of the Semantic Web that the difficulty of getting sufficient people to agree a sufficient breadth of vocabulary would forever stymie realization of the vision; despite being based in part on a false premise about the nature of the Semantic Web, this criticism has been hard to argue against compellingly.
I felt that this slide, coupled with the exposition of the Web's fractal nature (which for me echoed a WWW conference keynote given by Tim in Paris, just 10 years ago), went some way to providing a powerful basis for engaging with that criticism.
The slide itself ignored the size of ontology, which was probably the right thing to do for expository reasons, but when faced with the reality of developing an ontology I think it's an important consideration. In particular, I think that ontologies for very widespread use must be small if their cost is to be acceptable. (Or maybe this is a separate cost-of-use issue, where the presentation addressed just the cost of achieving agreement on a vocabulary.)
There was no specific mention of of ontologies vs folksonomies (roughly: formal vs informal vocabularies), but the Semantic Web is somewhat predicated on having formal relationships between terms.
5. Linking vocabularies
Related to the economics of ontologies, the "tube map" diagram was a very nice illustration of how a series of partial data overlaps can be merged (or "smushed" as the developers sometimes call it) to create a bigger picture.
The important message (following from the point about ontology economics) of this is that it is not necessary for everyone to agree the same vocabulary, as long as there is sufficient overlap between various communities of interest. Some very broad areas of interest correspond to long lines with many intersections with other lines; more specialized areas correspond to shorter tube lines.
6. "Do your bit"
For me, rhe "sound bite" of the talk was: "Do your bit. Others will do theirs".
I feel that developers of Web software often tackle more than they really need to, and consequently end up with less useful results.
http://www.oii.ox.ac.uk/collaboration/?rq=lectures/20060314 - OII announcement ofthe talk
http://www.w3.org/2006/Talks/0314-ox-tbl/ - slides for the talk
http://www.w3.org/2006/Talks/0314-ox-tbl/#(30) - the famous Semantic Web "layer cake", revisited. (An earlier version is , But see also .)
http://www.w3.org/2000/Talks/1206-xml2k-tbl/slide10-0.html - The "layer cake"
http://semtext.org/2004-02/slides/img4.html - Danny Ayer's version of the layer cake.
-- GrahamKlyne 2006-04-03 14:13:23