finally a bnode with a uri

Posts tagged with: sparqlets

ESWC 2009 Linked Data Dashboards

A first Paggr application went live during ESWC2009.
In case you missed the tweets or a local announcement: The first Paggr application went online a few days ago. This year's ESWC Technologies Team pushed things a little further, with RFID tracking during the event and extended conference data that includes detailed session and date/time information (kudos to Michael Hausenblas for RDFizing even PDFs).

Based on this dataset, we provided a conference explorer and stress-tested the "Dog Food" server while at it. The system survived, but I also learned a lot. We used about 50 RDF stores for the different public and user-specific dashboards, which basically worked nicely. However, rendering non-ugly resource summaries requires a bit of endpoint hammering, and some of the more complex path queries resulted in timeouts. Yesterday, I had to create a mirror from the data dump to route a couple of widgets through a replicated (ARC :-) endpoint. But then this is also one of the powerful possibilities that come with semantic web technologies. You can often switch or double the back-end repository in no time, and without any code changes. (And as all the Sparqlets are created in a web-based tool, I didn't even have to upload a changed configuration file. I simply tweaked a SPARQLScript parameter.)

Anyway, there are a couple of public dashboards, in case you'd like to give it a try (it's still an early version), I also embedded a short screencast below. The system is going to be moved to a DERI server when the conference is over, but the URIs and data will probably stay stable. (And no, it won't really work with IE yet.) More to come!

HQ version (quicktime, 110MB)

Paggr screencast: Linked Data Widget Builder

A screencast about Paggr's sparqlet builder.
Running an R&D-heavy agency in the current economical climate is pretty tough, but there are also a couple of new opportunities for these semantic solutions that help reduce costs and do things more efficiently. I'm finally starting to get project requests that include some form of compensation. Not much yet (all budgets seem to be very tight these days), but it's a start, and together with support from Susanne, I could now continue working on Paggr, semsol's Netvibes-like dashboard system for the growing web of Linked Data.

An article about Paggr will be in the next Nodalities Magazine, and the ESWC2009 technologies team is considering a custom system for attendees which is a great chance to maybe get other conference organizers interested. (I see much potential in a white-label offering, but a more mainstream-ish version for Web 2.0 data is still on my mind. Just have to focus on getting self-sustained first.)

Below is a short screencast that demonstrates a first version of the sparqlet (= semantic widget) builder. I've de-coupled sparqlet-serving from the dashboard system, so that I'll be able to open-source the infrastructure parts of Paggr more easily. Another change from the October prototype is the theme-ability of both dashboards and widget servers. Lots of sun, sky, and sea for ESWC ;-)

HQ version (quicktime, 120MB)

Sparqlets and more updates: new editors and sparqlets.
OK, this is hour 32 without sleep, so I probably can't write much. But blogging still seems to work as opposed to coding...

There are some new components on the eternal beta site. (Probably not only) the DERIans couldn't really get used to my generic tools, so I went back to the drawing board and more or less rewrote the whole stuff (less URIs and qnames, more "classical web site" structure). The new browsers are still based on a shared standard class, but I divided the "resources" section into 5 customized pages (events, projects, documents, people, software) and developed a set of resource-specific SPARQL queries to generate better tailored views. The filter forms are more or less fixed, but I already had a lot of fun with them (although noone seems to owe me beer..).

It took me some time to work around several issues, not only the bnode reference problem I mentioned, but I also needed a better index structure for the RDF store and basic inferencing capabilities in my query engine. I'm currently trying an experimental mechanism to pass OWLchestra-generated class trees to the SPARQL2SQL rewriter which then generates queries that allow me to retrieve conf:Workshops even when I only specified a type of ical:Vevent. Seems to work, not too sure about scalability issues, but I didn't like the idea of rdf:type 4ward-chaining either, so I'll stick to this approach for the moment.

There's a lot new (and hopefully more usable) stuff in the front-end. I've parameterized several of the framework's interface components and can now re-use what I call "Sparqlets": interactive page items that are based on SPARQL queries and JavaScript glue code. A "quick search" sparqlet in the upper right corner of the front page provides a suggest-as-you-type feature and allows you to easily search the whole RDF store. More sophisticated interaction possibilities are provided by the browsers, where you can now decide if a related resource description should be displayed in an inline sparqlet or as a separate summary. I've also integrated my unsmusher/resmusher in resource context menus, so that it's now possible to re-consolidate wrongly merged resource descriptions with a single mouse click, and without having to stop browsing.

but first: sleep.


No Posts found