I'm sitting in on the BioDASH demo, Eric Neumann is giving a broad strokes introduction to the semantic web, resources need metadata, basically if we can share and aggregate data and everything will be wonderful.
Now he's talking about semantic lenses, this I think is Haystack specific, he's talking about FOAF now. The slide he's using is the same one that was used in the bio-ontologies, for links see the comments here.
So he's pushing semantic browsers and lenses. Claims it is like CSS for semantic web: logic, aggregation, GUI, multiple perspectives of common knowledge. Of course this the the BioDASH Demo, and BioDASH is based on haystack so this is no surprise.
For those who don't know Haystack, it is a "semantic web" browser. I've tried out Haystack a number of times but it is the single most enormous java application I've ever seen, it does email, web browsing and anything else you can imagine. I've installed it a few times and never once managed to get it to start up and work in what I would consider a usable amount of time. Many times it just never started and hung the computer. It is after all a research project...
So Dennis Quan is now going through the user interface to Haystack, it is like a multi-window web-browser, has a graphical drawing kind of metaphor to drag around objects etc. (e.g. boxes and arrows where the boxes are the resources, the arrows are the predicates and so on).
Quite a few pointed technical questions at the moment: are the visual object resource and are the identified by URIs ? answer yes ? Does haystack support reification (he'll get back to it).
Runs through a demo of BioDASH, showing a number of boxes that represents compounds related to GSK3-Beta, also publications and other related information.
He's now showing a view of a pathway in Haystack, using data in BioPAX format. The BioPAX schema is a little complex, but they use a feature called lenses (like CSS for RDF) which formats the BioPAX data as a pathway (like cell designer). Basically simplifies or filters out some of the the nodes and data types in RDF to provide a user friendly view of RDF. This is important and kind of interesting... I should look-in this...
Many questions about the lenses (RDF styling feature), how is it implemented how does it work etc. Creation of RDF via the Haystack GUI ?
He is now showing the two views together, compounds and pathways, and says he will use Haystack to integrate these two views. Basically this is data integration using linked identifiers. EN mentions that the importance of this in terms of drug discovery, because all companies will have their own internal data models for this kind of thing. Using RDF can avoid data integration issues in this case (according to him).
Many questions follow.
Note: this is a REALLY canned demo because the URIs are known to be common between the two data-sets before they do the integration. Eric Neumann mentions this and makes the comment that in the real world we all need to follow standards and use LSIDs (completely avoiding the issue that in the real world, to date anyway, THIS DOES NOT OR HAS NOT WORKED). I will ask about this.
Okay demo is over. That was all, I think they are keen to have more discussion about this... no... they is now going on to talk about workspace personalization and haystack technology. Visualization technology in haystack: different views of the same data ? Next generation portals ? He claims this technology avoids a one size fits all for user interface. Easily generate an interface for chemists, biologists. This seems to trivialize the complexities user-interface design. This is why user-interface experts (e.g. Apple) should build GUIs and NOT engineers.
Only two slides on Haystack, now we are getting a technical overview of RDF. I won't write down all the content, just the interesting comments. Many questions about RDF, how is it represented, how it works, where do you get it, how do you convert data to RDF. Discussion of reification, what is it ? Statement about a statement etc.
Now we talk about LSIDs, this is good, they mention how it should work etc. Andrew Dalke is asking some fairly good questions, they all get down to the resolution issue, LSIDs are via SOAP ? What about URLs and content negotiation. So the discussion gets lively now (identification is a real issue in biology and the semantic web, for more on this search nodalpoint for LSID), how do I get metadata back from a LSID etc. Andrew asks: I can make up my OWN LSIDs that are compliant but return all kinds of stuff etc. how will that work with integration etc. I can only imagine that the BOF will get lively...
EN gives a summary, SWLS, defining LS needs, Show casing working examples, working groups, best practices. They are working on a charter, promoting LSIDs, BioDASH demo, identify issues. It's over, last talk of the day so off to the bar...
