Archive for November, 2006

2 papers worth reviewing from ISWC2006

Wednesday, November 29th, 2006 · Bijan Parsia

There were a fair number of interesting papers at ISWC, but I direct your attention to two:

Read the first to remind yourself that just because something is respected, respectible, and even cool doesn’t mean it’s useful (for your task). Read the second to remind yourself that formal semantics are good (or can be; done right!).

The first is a polemic, and a quite useful one. The second is a nice piece of theoretical computer science. I’d have to rate this ISWC as a good one since you can get at least one decent instance of each (i.e., a polemic and a bit of compsci theory), and these are better than merely decent.

The SPARQL paper won best paper (deservedly) and is serving as the basis of the new specification of the SPARQL language. Let me just say: Yay!

Pellet 1.4 is coming!

Wednesday, November 15th, 2006 · Evren Sirin

After a long pause, the next release of Pellet is now around the corner. When you defend your PhD thesis and welcome your first child during the same summer, software development takes a break!

There has been a number of changes since the latest release. The biggest change is the brand-new Web page Pellet has: http://pellet.owldl.com. We will be migrating everything (including the online demo, issue tracker, mailing list, etc.) from the Mindswap site to this new site. We currently have the release candidate for version 1.4 available for download at the new site.

The next release is adding a lot of new functionality as well as many bug fixes. The biggest update is the support for OWL 1.1. This version of Pellet supports all the features of OWL 1.1 except n-ary datatypes. So there is support for the long-awaited feature qualified cardinality restrictions, complex subproperty axioms (between a property and a property chain), local reflexivity restrictions; reflexive, irreflexive, symmetric, and anti-symmetric properties; disjoint properties, and user-defined inline datatypes. There is also an updated version of Swoop available at this site which supports OWL 1.1. The people seeing the demo at the 2nd OWLED workshop was quite enthusiastic about the new features.

In addition, the new version has features like query subsumption, axiom tracing for ABox assertions and a preliminary implementation of DL-safe rules. There is still some more work that needs to be done before the final release such as updating the documentation and examples. Well, until then, download the release candidate, play with it and let us know what you think (Yes, there is also a new mailing list for Pellet).

Cooperation, Competition, and Markets; Or: Why Expressivity Wars are Stupid

Monday, November 13th, 2006 · Kendall Clark

Attending OWLED last week gave me the chance to think about a part of the peculiar world of SemWeb
evangelism. In what follows I’ll describe what I think of as the peculiarity of that evangelism, and then
I’ll suggest a thought experiment that is relevant to ameliorating it.

Expressivity Wars are Stupid

The peculiarity is that SemWeb evangelists don’t seem to be able or willing to consistently distinguish
between the value of competition in a stable market and in a growing market. In short, competition is
useful in a stable market, but much less so in a growing market. In a stable market competition not only
helps buyers distinguish among relatively similar sellers, but it also helps spur innovation which would
otherwise not happen.

A growing market, however, is much more likely to grow if its sellers recognize the value of cooperation to
not only build the market, but to distinguish it, by cooperative action, from its competitors. All the
sellers in a growing market should work together to help buyers learn why they should abandon some other
market and take a risk on the growing one instead.

Every SemWeb evangelist should spend at least twice as much time explaining why (say) SQL and XML suck than why their pet SemWeb technology is better than some other SemWeb technology.

If we are all very lucky, the SemWeb is a growing market. There are some signs that it is, in fact, a growing market. I hope that it is, and I have made and continue to make business decisions based on that assumption.

My (rational) fear is that, in fact, the SemWeb market is not growing but is stable; and, thus, the
competition which I decry below isn’t stupid but is itself rational. But, again, I’m just not sure.

However, I am more sure that public fights about the utility of various representational formalisms—RDF
versus OWL; rules v. OWL; OWL Lite v. OWL DL; OWL Full v. OWL DL—are stupid, destructive, and
inappropriate.

They are stupid because they are really just proxies for a fight about which customers matter the most. But
in a growing market situation, they all matter to someone. (They are proxies because there’s no rational
reason to snipe and fight about representational expressivity; it’s a matter of scientific fact, and only
people ignorant of those facts fight about them directly.)

They are destructive because they form the basis for a reasonable perception of instability, and not the
kind that’s good.

They are inappropriate because they overstate the value of competition in (what we all hope is) a growing
market.

Expressivity Wars and the “Original Position”

Okay, so that’s my real point here: I don’t think it’s time to start competing in earnest yet; hence, I
think fights about representational expressivity do more harm than good right now. (Later on they’ll be
perfectly appropriate, rational, and helpful!)

But maybe you aren’t convinced. The rest of this piece is meant to convince you. Read on if you think you
might be convinced by more good reasons.

My attempt to convince you is basically a big analogy to one of the most important devices in contemporary
political philosophy: John Rawls’s Original Position.

Yeah, I know: what the hell is that, and how does it appy to the SemWeb? Glossing tons of details (when not
just getting them wrong…), here we go:

Working in the same social contractarian tradition as Hobbes, Locke, and Rousseau, Rawls wanted to give an
argument that rational agents would have to accept about what form of social arrangement would be the most
just (that is, would maximize social justice). So he cooked up a device called the Original Position in
which rational agents exist behind a Veil of Ignorance.

So imagine that you have no idea what kind of person with what kind of social setting or natural
advantages, etc. you might have. Behind the Veil of Ignorance, you don’t know if you are rich or poor or
famous or a good athlete or logically-inclined or work with your hands or a mid-level manager, or… You
don’t know anything about that complex constellation of social and personal details that determine what we
might call your “lot in life”.

Behind the Veil of Ignorance, what kind of social arrangements is it reasonable for a rational agent to
prefer? Rawls’s argued that those arrangements will be fairness maximizing because if you don’t know who
or what you are or where you come from or how much you have, then the only reasonable solution is to
arrange things in as fair a manner as is possible. In other words, if you don’t know whether you are rich
or poor, it makes sense to arrange things such that the gap between rich and poor is as small as possible.

Rawls put it like this:

  1. Each citizen is guaranteed a fully adequate scheme of basic liberties, which is compatible with the same scheme of liberties for all others;

  2. Social and economic inequalities must satisfy two conditions:

    • All offices and positions must be open to all under conditions of fair equality of opportunity;

    • Economic inequalities are only permitted insofar as they are to the greatest benefit of the least well off members of society.

Or: things have to be fair, except if some unfairness makes those who are worse off better off.

But what does this have to do with SemWeb evangelism and expressivity wars?

I suggest that in a growing market, there is so much opportunity, that is, uncertainty about outcomes,
that market agents effectively function behind a veil of ignorance. None of us really knows where the
golden gooses are. We have guesses and hunches and existing commitments, but we don’t really know much
for sure. In particular, since no one knows which customer’s problem—if any!—will be the killer SemWeb
app
, we don’t really know which expressivity regime is the most valuable. When low-expressivity people
argue for “just a bit more than SQL”, that can sound plausible. When full first-order logic people argue
for lots of expressivity, that can sound reasonable. And when, say, DL or rules or some other group
argues for their bits, they often sound reasonable too.

But, or so I suggest, none of us really knows. Which means that the only rational position is to
maximize the number and quality of the available options. A DL company may have to shift because of
unforeseen market realities to higher or lower expressivity formalisms. The rules-uber-alles group may
have to shift to DL for the same reason. Or we may all have to downshift to RDF plus a little bit of
OWL.

People sometimes act as if they know which will win (again, if any!), but that’s just various degrees of
bluff and bluster.

Since no one really knows for sure, I suggest that we should stop premature competition and public sniping
at the other guy’s best bet and maximize in the short and medium-term the degree to which we cooperate in
sending the message that our market is better than theirs. That is, that just about any of this
“semantic web stuff” is better than Yet Another SQL/PHP app, or Yet Another Database Silo, or Yet Another Half-Baked Java App, etc. I further suggest that this is in our
collective and individual self-interest for the same reason (by analogy, anyway) as Rawls’s arguments for
justice as fairness. In short,

Since you don’t know what technology will win, make sure you preserve as many potentially winning options, both individually and collectively, as is possible and consistent with your continued existence as a viable market participant.

Anyone who can’t do this because of their ideological commitments—that is, who’d rather not win at all
than win with the “wrong” technology—is just too ideological and inflexible to be of much good to anyone else.

An OWL 1.1 Working Group

Monday, November 13th, 2006 · Kendall Clark

The 2nd OWLED workshop has successfully come and gone; and the 2nd biggest news is that next year’s will be held in Innsbruck—which works fine, since Katie and I were planning to take our parents to Rome for a week at just about the same time, so it looks like a good deal of June will be spent in Europe. Woohoo!

The 1st biggest news is the increased momentum to do another version of OWL, which is being informally called “OWL 1.1”—though, as I suggested at OWLED, “1.1” or “2.0” or some other revision marker is a marketing decision and should be made in the future.

Yes, as with every W3C charter ever, there are some factions and proponents and opponents, etc., but this time it all seems relatively low-key and workable. I don’t hear any deal-breaker positions being taken yet that can’t be reasonably (if not happily! :)) accommodated. This is a good thing.

As I also tried to point out at OWLED, when you get 65 to 80 people (10 to 20 W3C-member organizations, plus probably 30 interested non-members) in a room and there are only 3 people who aren’t sure that another WG should be formed, while everyone else is sure, that’s called consensus.

Bijan’s last post has nice coverage of some of the technical details, but I wanted to use this post to just celebrate two nice facts: next June in Italy and Austria (yay!!); and OWL is showing all the signs of a vibrant technology on the move.

Not so alternate a view of OWL 1.1

Monday, November 13th, 2006 · Bijan Parsia

I’m back in lovely Manchester, but I’ve not recovered. Full postings and discussion of ISWC, OWLED, next steps for OWL, etc. shall have to wait.

However Jim Hendler has a blog post entitled “An alternative view for Owl 1.1?” The title is phrased as a question, so I think I’ll answer it: No, the view sketched in that post is really an alternate view. Or, at least, not a disjoint view.

To present the nub, here’s a quote:

In short, either as part of OWL 1.1 or opposed to it (depending on the poilitics) we should have some people working on a clean subset of OWL that is easy to learn, easy to support, and provides useful new functionality at the data schema/controlled vocabulary level — I think this would increase OWL use and uptake, and would provide a natural path from RDF to OWL through this middle layer – providing more room (and thus more money to be made) in the OWL space.

This is good advice….so good that it’s been part of the constallation of OWL 1.1 specs for quite some time, to wit: OWL 1.1 Web Ontology Language: Tractable Fragments. These include three specifically database oriented fragments, DLP, hornSHIQ and DL-Lite. (Note that there were papers at OWL on the hornSHIQ and DL Lite families.)

Some of these already explicitly address functionality Jim is after (e.g., see inverse functional properties in DL Lite). While DLP and hornSHIQ are closely associated with Datalog (i.e., being “Datalog” fragments of OWL), DL Lite makes novel use of databases (e.g., query answering by query expansion).

Are there more fragments? Of course! Considere Herman ter Horst’s paper, Completeness, decidability and complexity of entailment for RDF Schema and a semantic extension involving the OWL vocabulary.

An aside: “tractable” is really the wrong name for this set of fragments (and not only because from some perspectivies, they aren’t). “Interesting” is better (and I think there’s wide agreement that OWL Lite failed to be usefully interesting). Some things that make, to my eyes, a fragment interesting is that it has a different “natural” implementation technique, that it has low data complexity, and that it seems to have “natural” or “useful” expressivity. All the fragments in the Tractable Fragments document arguably have (and have been argued to have) these properties.

So, clearly, the idea that it would be helpful and valuable to the OWL community to “do OWL Lite right” or rather “do the species of OWL better” is a part of the current OWL 1.1 effort. Tweaks are, of course, welcome (so I’ll be interested to see what Jim and Ora come up with). And in the very very straw charter I jotted up, additional tractable fragments or refining the current set in other ways is clearly a big part of the work of the group.

So, seems like there is agreement, not disagreement.

P.S. I am extra confused by Jim’s “p.s.” where in he writes:

And I should mention now, early, and in public, that if the “charter” being circulated for OWL 1.1 were to come to a W3C I would oppose it on the grounds that moving away from the RDF syntax is a non-starter with me – but that’s a topic for some later blog…

Given that the current set of OWL 1.1 documents include a mapping to RDF graphs which is very much inline with the current mapping of OWL to RDF graphs. So it’s hard to see it as a “moving away”.

I’ll be interested in the future blog post explaining what’s up. Obviously, adding new, normative non-RDF syntaxes that are not discouraged as an exchange format means that RDF/XML’s status as the canonical and ubiquitous exchange format is threatened. But given the greater easy of mapping back and forth between the functional/abstract syntax and RDF graphs (i.e., there is a technical improvment in the mapping to eliminate or minimize the non-determinism of the mapping), I fail to feel the bite of this. There are many circumstances where being forced to work in or exchange RDF/XML is a real problem: e.g., parsing big documents, or making use of XML technology like W3C Schema directed editing come to mind.

P.P.S. Now I really am going to bed.