- From: Lee Feigenbaum <lee@thefigtrees.net>
- Date: Sat, 17 Apr 2010 09:12:08 -0400
- To: Alexander Johannesen <alexander.johannesen@gmail.com>
- CC: Danny Ayers <danny.ayers@gmail.com>, Pat Hayes <phayes@ihmc.us>, Semantic Web <semantic-web@w3.org>
On 4/17/2010 6:30 AM, Alexander Johannesen wrote: > Hola, > > Danny Ayers<danny.ayers@gmail.com> wrote: >>>> If we had compelling enough applications of the *data*, wouldn't we build >>>> the tools we need? >>> >>> Why? >> >> Because I want to know where the nearest kennels are, and when will be >> best to plant tomatoes. > > No, no; why is there some automatic notion that if our data is > compelling enough, the tools we need will be created? I'm not > questioning the need nor want for compelling applications, only the > assumption that once we get to stage 1, stage 2 will automatically > follow. We're all looking for that killer application, but perhaps > we're mistaking the killer app for techies for the killer app for the > real-world? *shrug* I suppose it's not an ironclad conclusion, but I'm a pretty big believer that if there is a compelling use of the data (not compelling data, per se) that we know about, someone will do the leg work to build that application. particularly if it's a monetizable application. >> Google do seem to have noticed that the hocus pocus (whether or not >> they call it RDF) has its place. > > I was more pointing to RDF being the culprit. When Google wants to buy > a few million bibliographic records, do they embrace MARC, MARC XML, > MODS, MADS, or RDF versions of the same? Nope, they create some simple > XML to represent the very basics they feel they need, and use that. > Same with most of the RDF data; silo mentality of the value of > datasets are incredibly hard to evaluate in the Linked Data world; you > have to take on good faith that the quality within is good enough for > whatever killer app you're writing. And quite often you only discover > lacks and poor data quality once you've gone down the path of > development for a while, never a pleasant journey. Are you expecting > killer apps based on data with faith-based quality control, and big > hurdles for evaluation of value? > >>> The Semantic Web was crafted on the potential of fixing problems a tad >>> bit better than what we already had that already fixed the problems, >> >> I disagree somewhat - would take me a while to find the exact quote, >> but Tim has stated words to the effect that the semweb can make >> problems previously considered impossible become a bit obvious. (A >> point with which I agree strongly). > > You are of course right, but all of that is theory. In practise we are > rehashing old problems in new ways. I guess what you're longing for is > the tipping point of going from solving those problems to solving new > ones. I think that's a good summary. Personally, I embrace the family of W3C Semantic Web technologies particularly because I find many of them to be a standardized form of what I'd otherwise consider best practices for solving problems anyway -- so I'm more in line with your original observation... I just happen to be happy enough to be solving problems a bit better than I would otherwise. (And for some classes of problems, I've seen "a bit" be big enough to be the tipping point between feasible and infeasible -- I'm not sure I've seen that with Linked Open Data though.) >>> so basically fixing a non-existent problem. It was also built on the >>> promise of reusable ontologies on top of data, and even though the >>> promise wasn't held the potential is still there, for sure. But we >>> haven't got the tools to deal with that part of it all that took us >>> (speaking in generic fuzzy terms here) by surprise; >> >> But we (in the affluent West at least) each have the hardware, >> software and connectivity to put us in the zone of making real use of >> this stuff. I still don't understand why we are so slow at making it >> so. > > Because we suck at coming up with good ideas, and even worse at > throwing something together to prove a point. If this stuff was easy, > we probably would see tons of it. But we don't, and I suspect that the > tooling sucks in a sense that it is hard for people in the real-world > to wrap their heads around them. SGML was brilliant, but hard to fully > grasp. And we know who's your generic markup daddy. > >> "informolasses" goes straight into my vocab, thanks. > > You heard it here first. :) > >> I suspect you're right about domain-specific tools, that reflects the >> human issues, the need to solve specific problems. >> While the Web of docs can be very generalist, I'm not so sure the Web >> of (linked) data will be useful in the same way, at least in the near >> term. >> For example, when I'm in gardening mode, I want a gardening >> application - that uses global data but within a locale filter. > > I have tons of similar problems. Even online tools I know how to use > and hack and exploit can sometimes draw up a blank. Like finding a > Guinea Pig breeder on the south coast of Sydney when you need one; 1) > there might not actually be any, or 2) there is no information about > them on the web to be crawled. The problem is not that they haven't > published their details in glorious Turtle. > > But is this stuff really the same problem as the Linked Data and lack > of killer apps, though? Good question. As Danny observed, the examples of compelling applications (I shy away from "killer app" simply because it implies to me that there is only one--unless you're a cat, I suppose) that have been mentioned so far all have to do with applying local geo data to a broader information base. Are there compelling uses of Linked Data that don't fall into that category? (Similar to the fact that all good Web 2.0 mashups involve a map, I suppose.) Anyway, just asking questions here, I don't know the answer(s). Lee >>> All this data and their weak relationships are great to play with, >>> though, and it might shape things to come, but to get the masses to do >>> something interesting with it you need to convince them that >>> "ontology" is even a word that deserves a place in our daily >>> languages. (And don't tell me linked data doesn't need ontologies; a >>> kick in the shin if you do) Tough call, I'd say. If you say to them >>> "model", they immediately reach for Toad or some RDBMS thingy. If you >>> say "triplet" or, even worse, "tuple", they might think you're talking >>> about raising kids. >> >> Kick me in the shin - ontologies are no more and no less than shared >> vocabularies through which we can communicate. > > I can't kick you in the shin based on faulty reasoning or > understanding of what I admittedly poorly wrote. :) The point was that > Linked Data uses ontologies because, like you say, they're shared > vocabularies. Not the most complex vocabularies, of course, but > vocabularies or ontologies nevertheless. I doubt interchanging > "vocabulary" with "ontology" has the slightest effect on people's > understanding of how these things fit together, and *especially* not > the potential therein. > > What I don't understand is that people have no problem understanding > names of elements in an XML schema, and link that and its data content > to records or fields in a database (which is a fuzzy undertaking when > you get right down to it), but have huge problems taking a triplet or > two and doing the same. There seem to be some cognitive mismatch > happening when you introduce the tiniest third directional signifier. > It's puzzling. Is the human brain too capable of doing one-to-one > mapping that it fails our attempts at many-to-any? > >>> In other words, the technology, its promises and potential means >>> *nothing* when a small paradigm shift is needed. >> >> Despite my negative comments recently, I do think that paradigm shift >> is happening. > > Where and how? > > > Regards, > > Alex
Received on Saturday, 17 April 2010 13:12:48 UTC