Freebase Parallax: Yummy UI … But One Query

The unfortunately named and well funded Freebase, about whom I’ve previously blogged here, recently released a nice UI called Parallax. It’s the closest thing I’ve seen to a “BI-style” navigational query interface against “unstructured” data, and worth a look.


  • I quoted “BI-style” to remind everyone that search vendors didn’t invent multi-dimensional navigation as they seem to believe. The whole notion of facets (i.e., dimensions) goes back to OLAP vendors and to the EIS vendors who preceded them. Faceted navigation is a wonderful thing to do against content (and MarkLogic features it), but let’s remember who is immitating whom. Put differently, Endeca is PowerPlay for paragraphs.
  • I quoted “unstructured” because, despite appearing to have document entries, Freebase is anything but unstructured. It’s a dictionary-driven, metadata-driven Wikipedia. Concrete example: when you’re dealing with the San Carlos airport entry, Freebase “knows” about airports, for example, that they have runaways, which in turn have lengths, widths, and orientations.

The video does a great job of pointing out the difficulty in answering a basic question (“which schools did the children of Republican Presidents attend”) using either Google or Wikipedia. It also shows that the question can be answered navigationally rather easily using Parallax. That’s nice.

But consider this: in a reasonably well structured XML database, you can answer that question in a single XQuery. (See below the video.) Now, I’m not trashing sexy UIs or the need for end-user tools to query data. (Given my 9 years at Business Objects, I’d be the last person to do that.)

But I am saying that delivering end-user analytics requires two things: a nice front-end with a powerful DBMS behind it. And if data/content source for those analytics is, or should be, XML, then you know which DBMS I’d recommend.

Freebase Parallax: A new way to browse and explore data from David Huynh on Vimeo.

And here’s the single XQuery, courtesy of Kelly Stirman:

let $presidents := /person[@type=”US President”]/name
for $child in /person
where $child/relationship/@type=”parent” and $child/relationship/name=$presidents
order by $child/date-of-birth ascending

3 responses to “Freebase Parallax: Yummy UI … But One Query

  1. And in SPARQL the query would look something like:SELECT ?president ?child { ?president a wp:President ; wp:child ?child .}And I’m sure there’s something very similar in MQL, Metaweb’s query language. Parallax is really about the user interface atop a paradigm of set-to-set navigation; I don’t think it should be taken as indicative of the lack of a powerful DBMS behind the curtain.Lee

  2. In fairness, my example SPARQL query was not as complete as the example XQuery query. The equivalent query would be something like:SELECT ?name ?dob ?school { ?prez a wp:President ; wp:child ?child . ?child foaf:name ?name ; wp:date-of-birth ?dob ; wp:school [ wp:schoolName ?schoolName ] .} ORDER BY ?dob(The formatting on this will probably be horrible. Apologies.)

  3. C. M. Sperberg-McQueen

    Good points.But “the whole notion of facets (i.e. dimensions)” surely goes back further than OLAP and EIS. It’s the central notion of S. R. Ranganathan’s Colon Classification, which he began working on in the 1920s and which appears to have been first published in 1933. And, of course, Ranganathan formulated the notion of facets as a way to crystallize and make explicit some important regularities in the construction of the Dewey system.All the more reason, of course, to agree that it was not invented yesterday, and not invented by search vendors.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.