MishMash

Cameron Neylon, Capturing process

Just been at a seminar organised by the Comp Sci people here in Newcastle. Cameron Neylon, currently of the Science and Technology Facilities Council on:

Capturing process: The promise and challenges of connecting experimental records to the semantic web

The promise of the semantic web in its broadest terms to integrate structured data, and in particular scientific data, is breathtaking. Yet our ability to deliver on that promise remains severely limited. On one side, battles are fought over vocabularies and structured description languages, while on the other computer scientists argue over design and architecture. In the middle the experimental scientist is usually left bemused as to what all the fuss is about and how it can have any relevance to their research, which is naturally far too specialised and ground-breaking to fit into any framework that an outsider could ever provide. The challenge lies in finding a pathway that moves towards satisfying the needs and interests of these disparate groups while providing both value and a usable interface to the end user.

We have adopted a blog based approach to capturing the record of research in an experimental laboratory. The use of a free form framework along with templates encourages the user to structure the description of their work in a way that matches their needs. Cycles of template design and optimisation lead to structures with strong similarities, and occasional important differences, with designed ontologies and vocabularies. The challenge for the future lies in determining how best to guide the actions of users towards the use of appropriate controlled vocabularies and structured descriptions, while using their choices to inform the design of these systems. The potential for Google Wave to provide a framework to enable local structuring as well as integration with global structured descriptions will be explored.

Some interesting ideas, as abused by my memory:

• People tell stories, they don't organise stuff the way it needs to be organised for machines to process. A big part of the disconnect between scientists and the folks building "e-science" tools is that these tools don't help people tell stories.
• So we need to start capturing structure through storytelling. That is, allow people to work in a natural way and pick up the structure from what they do.
• You can create a virtuous circle between metadata, which a user would normally not bother filling in or fill in incorrectly, and user interface. Arrange the UI so that the metadata populates the UI. The user wants that UI to make their life easier, and they start to notice that getting the metadata right is necessary.
• (Lab) instruments as bloggers. Have them push results straight into the blog. Take the human out of the loop (I might qualify, as I think was intended: where they add no value).

Cameron has had early access to Google Wave, it seems, and is impressed with the potential. He had some cool examples of Wave "robots" going off and filling in extra information, drawing graphs, creating links, tidying up references.

He was less impressed by the Google client: too much stuff going on in a single UI, too non-standard. He also expressed concern about users understanding the model. It's the eternal problem with geeks developing stuff for "normal people" again. We're geeks because we do abstract with ease.

There was a lot that I recognise, and some new thoughts as well. I'm all about computational analysis in engineering decision making rather than lab-based science, so there are some differences, but there's a lot of overlap too. In theory we should have an easier time here since so much of our "stuff" comes in the form of information artefacts, and our "experiments" take place in silicon. There are instruments, and the process of getting data from them into our analysis is currently extremely painful, but most of that is "upstream" of our analysis.

I liked the picture of an lab process generating artefacts and narrative, and then a separate process of arranging these things into a different narrative for publication. The story we tell when we want to explain our results to others isn't the story of the trials and tribulations of getting to those results, its a story that gets the reader as smoothly as possible from their current knowledge to where we want them to be. That's true of engineering analysis as of science. And in both cases, we'd be in a much better place if, given that latter story, we could easily investigate the artefacts referred to: look at them, do our own visualisations or further processing, and follow the trail back to see exactly how they were generated.

Digital information is different

This is beautifully done, and hits a few nails on the head:

Merlin has a few worthwhile comments on the strengths of paper. Also missing, from my point of view, is the fact that a lot of what we do with computers now (pretty much everything that's off the web, for a start) needlessly recreates the limitations of paper and adds a few new problems for good measure. Now where did I put that file ...

Tim Bray interview at ACM Queue

Another interesting interview over at ACM Queue: Tim Bray, mostly on XML and RDF.

The XML serialization of RDF is horrible;   it’s a botched job.

Tim Bray

Amen. Shame that.

[RDF] quickly became mixed up with a whole bunch of classic KR (knowledge representation) people who wanted to go refight the AI wars of the ’80s. … You know, KR didn’t suddenly become easy just because it’s got pointy brackets. Doug Lenat has been off working in the desert on that for decades and nobody has ever made a buck on it yet, as far as I know.

Tim Bray

There surely has been a buck made on KR, albeit not on Lenat's work (the Cyc project). Representing a significant amount of human knowledge formally is way too expensive to give a positive ROI, perhaps, but formal logic based languages have great promise in far more situations than they are currently used in. The current state of RDF is that it is too limited for it to be of much help there, and this is a pretty unsurprising result of its history. What was supposed to be a language for encoding metadata has been dragged into being a KR language, but it remains a seriously limited one. In the mean time it attracts a load of funding at the expense of other work.

Limitations of RDF as a KR language are limitations, too, as a metadata framework. After all, what use is a metadata framework intended for deployment on the Internet that can't represent provenance of data? This is being worked on, of course, but if RDF had been based on an existing KR language which supported multiple contexts, it wouldn't be an issue now.

Taxonomy, folksonomy, and value

Let's try again.

Shelley Powers has written a good summary of some arguments about the relative merits of tagging and more formal metadata. Nice pics, too.

Shelley is worried that the rise of "cheap" metadata, in particular tags, will inhibit development of tools to make "real" metadata available to the masses.

Theory: the value of terms in any metadata system used in an uncontrolled way tends to zero with time.

That's just as true for rich, formal so-called “ontologies” as it is for tag soup. There is a difference, however. Tags don't lie about their status. We know, right off, that they're just terms which people have attached to things, for some reason. We know that we don't know what sense the term is meant in, of the undoubtedly many possible. We know that we don't know the intention of the tagger in tagging. The spam problem arises from this: spammers are applying tags in order that pictures show up in certain places, not because they think those tags are relevant. The business of tagging flickr photos with the "offensive" tag is another case in point, where the intention is now to stop items showing up in certain places, and really has nothing to do with offensiveness.

Problematic, perhaps, but not unique to tagging. Exactly the same is true of uncontrolled use of any classification system, however good the tools are which provide access to it. Things will be tagged in a way which is inconsistent with the intended use of any system, either through ignorance or intent.

The rel="nofollow" tag which Google has announced it will use to control the inclusion of links in its PageRank calculations is indeed a nice demonstrator of an important point here. It's meaning is very well defined: it is grounded not in human understanding, but in machine behaviour. Nofollow can't be spammed: either Google honours it, or they don't. The web now has a mechanism for systematically witholding page rank. What it does with it is out of Google's control.

Point is, that some terms have meaning in this, mechanical way, like those which make up a programming language. Others have meaning in a fluid, human way. The latter will always be as fluid as they are human; no fancy tools will change that.

Fancy tools to make use of formal taxonomies easier would be great, of course. Very useful for those who care enough to put in the effort to understand, and make sure they stick to, the formal intended meanings of terms. But however easy that is made, it will always be hard, and people will always get it wrong. And as people get it wrong, the extensional definition of the term becomes less precise, and the value of the term tends to zero. You might slow the process down, but you can't, simply can't, stop it.

And, of course, because the effort involved in understanding and carefully applying the terms of a formal taxonomy is relatively high — independently of tools, which can reduce the accidental difficulties but cannot attack the essence — it is just not useful in the sort of "fire and forget" systems like del.icio.us, where any effort is too much.

In the comments to Shelley's post, Nick Sweeney writes:

I’m a bit of a Saussurean about this, in that I think that taxonomy (or ontology, depending upon your disciplinary point of origin) is crystallised/calcified folksonomy. Authorised folksonomy, if you like.

Right on! Those two terms, “crystallised” and “calcified” are a really useful duo. The one has connotations of order, beauty, and value. The other of dull rigidity. Well designed formal taxonomies have the first of those in abundance. Both have the second, and you can't avoid it. A taxonomy can only ever be an encoding of one incomplete and imperfect view of the world.

Shelley:

I agree with Clay that the semantic web is going to be built ‘by the people’, but it won’t be built on chaos. In other words, 100 monkeys typing long enough will NOT write Shakespeare; nor will a 100 million people randomly forming associations create the semantic web.

100 monkeys indeed will not write Shakespeare, but 100 Shakespeares didn't write Shakespeare either. Shakespeare did. The situation we have is that of 100 million people (or, at any rate, some large and growing number) who need tools to organise stuff, and to help them find stuff. 100 million people who will go right on randomly forming associations, whether the results pretend to be anything other than randomly formed associations or not.

We're dealing here not with the formal space of so-called ontologies and logic, but with the messy, human space of language and meaning. The processes involved are, essentially social. Tools which deny this will fail.

Shelley:

Clay believes that ultimately ontologies will fall to folkonomies, as the latter gain rapid acceptance because of their low cost and ease of use; I believe that ultimately interest in folksonomies will go the way of most memes, in that they’re fun to play with, but eventually we want something that won’t splinter, crack, and stumble the very first day it’s released.

When you design a building to ride out earthquakes with minimal damage, you build in flexibility. Of course, the clever part is in deciding where and how to build in that flexibility. Tagging provids an abundance of flexibility, so much that the building can barely stand up.

But in some ways it works. Delicious is already very useful, on an individual basis, and has some use at a social scale. I follow what other people post about Lisp, Scheme, and Smalltalk for example. These are well defined terms, referring to programming languages rather than highly ambiguous and mis- or differently interpretable concepts. With less well defined concepts, the problems lie as much in the nature of the concepts as in the tagging approach.

Simply aggregating tags a la Technorati as it stands doesn't show a great deal of promise, but it is possible to envisage tools which show tag clouds and, by presenting these to the user, encourage convergence around particular tags. The results will never be complete, never perfect, but always changing, and sometimes useful.

When you have a Web-load of people, things happen from the bottom up. That's the way the Web works.

Jailed for using a nonstandard browser?

How's this for two sides of the same story?

It matters because it has a URL

Blogs matter because they have a URL. They have a permanent web address. So often on the Internet, or in technology, but particularly on the Internet because of the social effects, a little tiny thing makes a huge difference. So this is a little thing, that weblogs have a permanent web address, because what that does for them is give them permanence, so we can keep going back to them and learn about the person.

Dave Weinberger

The importance of the humble URI has been coming up on and off recently. The quote above is from an after dinner speech Dave Weinberger did at the Blogging, Journalism & Credibility conference, which I had a listen to on my walk home (the walk home's too short — I had to listen to the end while I was making a cup of tea and some toast). In a discussion on the geo-reasoning mailing list, I commented regarding RDF that,

This could have been avoided if the TimBL and co had adopted a well establish knowledge representation language and "webified" it by using URIs for terms, rather than starting over.

To which Malcolm McClure replied,

I think that hypertext links, etc are very much a secondary issue, …

Aaargh. No! URIs — for all their faults — are very much the primary issue. And then there are all those broken web sites. The UK notional rail enquiries site has much more pressing problems to address (it's just broken) but if it worked at all it would still be a nuisance that you can't bookmark a train. Sure, it'd take some thought to work out how to encode a journey in a URI, but it'd be well worth the effort in the usability increase. How do we get this message over: if it doesn't have a URI, it's not on the web. If it matters, give it a URI.

Update (2005-02-01 (Oh heavens, it's February already!)): Make them nice URIs as well, of course. It's not hard. I found ISAPIrewrite, which has a free limited version, when we were sorting out the School's web site to fix this very problem recently. Apache does full regex-based rewriting out of the box, of course.

Amoral Adsense

Over at Bill de hÓra's weblog there's a post about a thesis on distributing RESTian web services. In itself worth linking to for future reference, but what caught my eye was the ads Google has placed at the top.

Custom Dissertations! it says. Custom Unpublished Dissertations in the UK - Quality Guaranteed.

UK University Assignments!

An impressive demonstration of Google's adsense ad placement algorithm, to be sure …

Artificial Intelligence? Pah.

Long ago Dad introduced me to a quote, which he believes to have been first uttered by a head of the AI group at the University of Edinburgh.

Artificial it may be, Intelligence it most certainly isn't. [Source unknown]

I have in the past found Google to be quite effective at locating sources of quotes for me. In this case it is of no assistance whatsoever. If anyone has any leads on this, I'd love to know. If not, you heard it here first.

Spread your wings and fly, little meme.

timeanddate.com

I've just come across timeanddate.com. An essential for anyone who has to arrange virtual meetings with people in different countries, like this.

Another David Harvey

Here's an interesting namesake:

I've written articles on Smalltalk, Prolog and Scheme, and reviewed books for the BCS OOPS newsletter and for SIGS European journal Object Expert. I've spoken on C++, and presented workshops on software architecture, architectural styles, patterns and development practice and culture at numerous conferences and events, including several OT conferences, JACC, Object Expo Europe, Unicom, Software Architecture 2000, and JSIG. [David Harvey]

And a good demonstration of why I call myself Hamish. This came up from GoogleAlert. I have alerts for "David Harvey" and for "Hamish Harvey" among others. David Harvey is always full of rubbish, and lots of references to the David Harvey who wrote "The Condition of Postmodernity". Hamish Harvey surprisingly often has items which refer to me (mostly written by me, of course).