green_amber

Feel free to pass it on!

From:

bugshaw.livejournal.com

My first stab at this is: data is relevant in multiple contexts - joining ontologies in a cross-linked web structure rather than a tree hierarchy might better model the way the data is found and used. Perhaps monitor the popularity of various cross-links, to provide insight into the most common inter-ontological relationships?

[This is off the top of my head without a decent technical knowledge of the Semantic Web or ontology. I'd be interested to learn more ;-)]

From:

voidampersand.livejournal.com

Yes, I think the simple answer is that a single ontology is unlikely to ever exist, so if the semantic web is to work, it needs to work with multiple ontologies. This is not perfect, but it's better than the current state of the non-semantic web.

From:

Thanks. me too :-)

From:

Does the Semantic Web imply the need for one ontology to rule them all?

God no! (Sorry, just got back from lecture (obv must check LJ first thing, will answer more fully on my journal.)

Having spent most of the last week updating an ontology based on a schema that recently changed from v1.20 to 1.30, I quake in fear at the thought of one big superontology driving an application that should be implementable in multiple ways. Or I could just be hallucinating with all the Pacman yellow class bubbles in Protege. But thank you for asking.

From:

the-gardener.livejournal.com

acquired a rather wonderful ecru linen transparent coat thing in Per Una to go over a brown/cream sundress

I thought Per Una was an M&S brand, not a store in its own right? Or does it differ north of the border?

From:

It is but we don't mention it :-) In Edinburgh per Una has got so popular it now takes most of the ground floor of ladies M & S..

From:

pisica.livejournal.com

Hee. You're the second lecturer I know of to be quietly marking exams - and since the first one is very pro-union, I expect there are plenty more....

From:

bohemiancoast.livejournal.com

My lecturer eventually caved too. I passed! hurrah!

Looking forward to seeing you tomorrow.

From:

Congrats!!!! WE are not caving though :-)
On 2pm train is idea - be with you c 7.30/8?

From:

andrewducker

Does the Semantic Web imply the need for one ontology to rule them all?

Bwahahahahahaha.

There is no such thing. Tisn't possible.

From:

Home already!
Public transport exceeds expectations yet again..

From:

andrewducker

Heh. It only takes about half an hour to get me home from there - 10 minutes to Princes Street and twenty from there to my nearest stop. Oh, and 7 minutes walk down.

Now for sleep!

From:

The Semantic Web most certainly does not require a single ontology, not least because an ontology is an artifact designed with some purpose in mind. Ontologies are naturally both domain- and task-dependent, and it's really only the anti-SW crowd who claim otherwise, that the SW is some Diderot-esque fantasy of a singular, universal, self-consistent representation of human knowledge. The rest of us (being knowledge engineers, logicians and SW practitioners) know how hard it is to elicit sufficient knowledge from a group of experts who barely agree between themselves.

The tax law example you give is an example of a multi-perspective ontology. We have a system at Southampton (remind me to show it to you when you're down) that supports the triple assessment protocol for breast cancer treatment; a patient is seen by a clinician, a radiographer and a histopathologist, all of whom have different viewpoints on the same subject, and use different terminology. Where the radiographer might talk about microcalcification visible in X-rays, the histopathologist would be looking at cellular abnormalities, and the clinician a lump in breast tissue, but these are different manifestations of the tumour.

Working with all three types of expert, we developed viewpoint-specific ontologies and provide the tools to map between these ontologies; there's some minimal agreement at a high level, so there's a top-level ontology which describes a framework into which the individual perspectives fit. You could develop the sub ontologies in isolation, but you'd be making the integration harder for no good reason (better to plan ahead).

From:

That's .. mindblowing for me as a lawyer. meets a lot of the problems B and I have been talking about. Wanna see. BTW, seem to have lost your email address - can you mail me? Are we going to meet June 5 and if so when would suit? I also want to mail you latest from prospective PhD student..

From:

bugshaw.livejournal.com

That's .. mindblowing for me as a lawyer.

ISTM that in Law, you learn the ontology and then work within that framework. Given a new case, you investigate to see how it best fits, what existing law or judgements apply. You may not stick it in ad-hoc, or create a new category. Because, you know, it's The Law, it is not optional, and you do have to work within it. Ignorance is no excuse. "Common sense" comes second where Law exists (I'm thinking of the common idea that "if my parents give me their house then live for 7 years, they can still live in it but I won't have to pay Inheritance Tax").

In (most) other parts of life, things work the other way around, or maintain a balance between working within ontological strictures and growing new systems.

From:

Have mailed you. 5th June is probably not going to be possible, unfortunately (research group awayday, with threats of unspecified sanctions for non-attendance), but I'm free all day on the 6th at present.

From: (Anonymous)

we developed viewpoint-specific ontologies and provide the tools to map between these ontologies

Do you have any online documentation or papers that you could point me to? That would be fantastic. I'm working with ontologies in space science.

From:

Easterbunny?

From:

D'oh! Yes. Login::brain interface failure.

From:

Our work system seems to log me out every other am at the moment. hence first post tending to be anon :(

From:

The project is called MIAKT - there's more information on the website, including papers, videos, and some OWL ontologies.

From:

Thanks! There's a roundup of the virtual observatory schema -> ontology work here, access control stuff here, and the big universe one here.

From:

Ah, right. Tony Hey gave a keynote on eScience yesterday (I'm at WWW2006 in Edinburgh), and he mentioned this in passing.

From:

drdoug.livejournal.com

Does the Semantic Web imply the need for one ontology to rule them all?

As others have said, no it doesn't. Which is just as well, since most people (me included) regard such a thing as impossible.

Can all 3 start working independently on their ontologies and eventually usefully share data in applications, or do they all have to agree an ontology at scratch (which just ain't going to happen?)

There's no magic in ontology engineering. If the groups don't have a shared understanding of what their data means, they're not going to be able to meaningfully exchange it, and there it is. They can come to that shared understanding now, or later. This is a trade-off: if they do it later, it'll be a lot more work to make the data interoperable; if they do it now, it'll be harder since they won't understand how to describe their own data well and they won't agree.

(BTW, focusing on the ontology as the main aim is almost always a mistake. The focus should be on what they want the system to be able to do. What do German, UK and US tax lawyers want from a common system? If you don't have a good answer to that you've no business spending all that time and money developing an ontology, because it'll certainly be wasted. And IMO Semantic Web technologies are not necessarily the best way to meet the as-yet-unarticulated needs of these disparate tax lawyers.)

From:

That's interesting. I thought the whole point of the Semantic Web project was to get data out there descibed in ways which could then be appropriated for multiple applications. In other words, start with the data/ontology, think of the functionality later?

From:

ah, I am learning better - see

nmg above and my reply..

From:

focusing on the ontology as the main aim is almost always a mistake.

I'm not sure I agree with this. The people in my workgroup (loosely pulled together as "intelligent resource discovery") are all approaching ontologies from a different angle. One guy has a specfic application in mind for which he has developed a small, lean mean ontology. He gets major, "Wow, you made it do something useful!" stars. There's a guy building one from scratch that describes the universe. He's having to wade through a lot of domain knowledge disputes (is "wolf rayet" a star or a galaxy). I've been translating 2 community standard XML schemas into ontologies because I want to see if I can do something with a .owl format and its implied relationship / logic that can't be done with the .xsd formats we've had knocking around for 3 years. I have another 3 months to demonstrate that the solution-first-problem-second approach has a point; after that time it is entirely possible that guy #1 will be the only one with funding and they will find a different way for me to ~~gouge the taxpayer~~ make a useful contribution to space. If I can't build ontologies I want to bake space cookies.

From:

Funny. I want to know about this stuff because it might enliven my life - not because I think law really needs it. I am such a fraud :-)
Is anyone using an ontogy/SW application for an actual working practical project yet?

From:

When I had a look last summer, it's largely medical researchers, particularly in genetics and cancer research. Links to look for include Gene Ontology, Oncology Ontology, and the Open Biomedical Ontology Database.

The US military is testing out ontologies to automate battlefield decisions, but I don't have any links to the talks I heard last summer.

From:

re: the US work, this is all linked in with what they're calling Network-Centric Warfare and Battlespace Management, which they've been working on for at least the last five years. The OWL language developed from DAML (the DARPA Agent Markup Language) which began in 2000, and that program has direct antecedents within US defense research going back to the early-to-mid 1990s.

The UK has similar programmes under the name of Network-Enabled Capabilities, which also mesh with NATO-wide efforts to revise the LC2IEDM datamodel vocabulary that are widely used in military planning systems.

From:

drdoug.livejournal.com

Oh, sure - in research terms we still have a lot to learn about how to use ontologies and the Semantic Web, so if your success criterion is "What will push back the boundaries of human knowledge?" or more venially "What will get me more grants?" then you're likely to come to a different decision about ontology-first than if it's "What will get me a system that is useful in practice?".

From:

Cool - are any/all of these ontologies available online?

Another of the projects I work on at Southampton is a collaboration with our chemists, who are investigating ontologies and the SW simply because they've realised the expressive limitations of XML vocabularies such as CML.

From:

BTW, focusing on the ontology as the main aim is almost always a mistake. The focus should be on what they want the system to be able to do. What do German, UK and US tax lawyers want from a common system? If you don't have a good answer to that you've no business spending all that time and money developing an ontology, because it'll certainly be wasted.

I agree, ontologies for production systems should be designed with a specific task in mind, rather than being thrown together with no attitudes to future use. Having said that, the process of designing an ontology without such a tight set of use constraints can be a valuable exercise, because it allows the knowledge engineer to get some idea of the scope of a domain.

IMO Semantic Web technologies are not necessarily the best way to meet the as-yet-unarticulated needs of these disparate tax lawyers.

I'm not entirely sure that I'd agree with this (but then I would say that, given that I was on the working group that designed OWL). True, there are other KR paradigms (such as frame logics and datalog-based solutions) and the description logics on which OWL is based are conceptually difficult, but there's a extensive body of robust editors, reasoners and other tools that support SW technologies.

From:

drdoug.livejournal.com

I have great hopes of getting most my essays marked on the train. ( don't tell the AUT )

Do you also have great hopes of passing up any extra salary increase secured by your more solidarity-minded colleagues? *grin*

From:

Snerk. Exactly what I've been asking my non-AUT member colleagues.

From: