Slashdot Mirror

See my post here:

http://news.slashdot.org/comments.pl?sid=2412564&cid=37307402

Or enjoy reading through things like this, of which Google searches will turn up many:

http://www.edbott.com/weblog/2005/07/andrew-orlowski-is-a-hack/

http://paulfwalsh.com/why-andrew-orlowski-from-the-register-is-a-twat/

http://www.texttechnologies.com/2007/03/26/andrew-orlowski-berners-lee-spam-semantic-web/

http://blogs.computerworld.com/16711/why_andrew_orlowski_is_wrong_about_net_neutrality

http://ktetch.blogspot.com/2011/05/andrew-orlowski-drunk-unethical-or-just.html

But ignoring Andrew Orlowski there's countless issues with their other authors too. Lewis Page is more reasonable in allowing dissenting comment in response to his articles, but his articles are time and time again completely ignorant. He for example often criticises British defence projects citing American options as being much cheaper by pure monetary, but despite having it pointed out to him time and time again he fails to realise that a $10bn UK defence project for say, some new helicopters is still cheaper than buying the helicopters for $8bn from the US, when the UK project brings back $5bn in eventual tax, whereas the $8bn US project it's just money straight out the British economy.

Another example is the Eurofighter typhoons ground attack capabilities- he constantly derides the project because it wont have proper air to ground capabilities until 2020, but he's wrong because it wont have proper bombing capabilities until then- it's had Brimstone missiles added to it throughout this year. He ignores AGMs and focusses on bombing capability and then extrapolates that to say it can't do air to ground at all until 2020. This is complete and utter outright FUD.

He's similarly criticised the armament of Type 45 destroyers, claiming they only have two weapons or similarly, but a quick look on the Royal Navy's own website and the specs of the ships confirmed that yet again, he's completely wrong.

You can see this pattern with most of their staff- their articles are just often outright false. Where they're not false, they completely miss fundamental points. Where they don't miss fundamental points, they just outright lie.

So that's really why they have the reputation- they're just too agenda based. Their writers all vehemently pursue their own political agendas without care for facts, without care for reason, and worst of all- without care for the truth. That's not journalism, that's propaganda.

Big companies sometimes really do want to be "made" to do the right thing, so that they can do it in lockstep without losing competitively to their less scrupulous rivals. Whether it's bribery or pollution or safety or censorship -- whether the motivation is simple economics or avoiding later political backlash or actual morality -- it's happening more and more.

My links to http://www.texttechnologies.com/ weren't just of the shameless plug variety; I really have written more there arguing the case than can realistically be repeated here. That said, if I have to choose between them I lean strongly to what you're calling a "rigid" ontology, because if the ontology is just automagically generated from each corpus, one can't build any kind of application on top of it. But what I really want is a combo; think of it as 100% of the benefits of a "rigid" ontology, plus 50% of the benefits of a fluid ontology as well.

Anyhow, to make any kind of "rigid" ontology work, it clearly has to be much more easily extensible (both by hand and as the result of automatic clustering) than ontologies are today. That's the challenge for the new product category I'm calling for.

And by the way, relying on authors to tag their documents is a non-starter. You can't even get Wall Street analysts' reports, which go through a double round of editing so as to comply with SEC regulations, accurately tagged according to the stocks mentioned in them. And that was true even BEFORE the online era, when publishing cycles were several days long. Really. It boggles the mind.

Corporate search technology isn't good enough. And it's not going to get good enough until it's informed by much better ontologies/taxonomies. What's more, when the required ontology/taxonomy engines and business processes finally exist, they won't just be used for search; they'll also be used for text mining, knowledge extraction, speech recognition, and so on. So far as I can tell, none of the big generalists or small linguistics specialist companies understand this point, or else the ones that do are really small and lack the resources to do anything about it.

The first company to break that logjam will be a huge winner, with a market opportunity comparable at least to that of, say, app servers. Unless, of course, the whole thing is just open-sourced.

I've written up most of that argument in http://www.texttechnologies.com/2005/12/11/the-tex t-technologies-market-3-heres-whats-missing/ and related articles. The big missing piece (about the open source alternative) is coming soon, on the same site, and perhaps in my next Computerworld column as well.

One thing in Google's favor, however, is their internal use of knowledge extraction. They seem to really be ahead of the competition in that regard.

Meanwhile, I've got to say -- search is one of the areas where Microsoft has been saber-rattling for a long time, to little effect. Just a couple of quick examples of what I mean:

1. In 1997, I was at the Verity user's conference, and a Microsoft guy there told me how Microsoft would soon be in the business.

2. A few years ago, a woman emailed me and told me she'd just joined Microsoft, and was personally writing all of their web search algorithms.

Corporate search technology isn't good enough. And it's not going to get good enough until it's informed by much better ontologies/taxonomies. What's more, when the required ontology/taxonomy engines and business processes finally exist, they won't just be used for search; they'll also be used for text mining, knowledge extraction, speech recognition, and so on. So far as I can tell, none of the big generalists or small linguistics specialist companies understand this point, or else the ones that do are really small and lack the resources to do anything about it.

The first company to break that logjam will be a huge winner, with a market opportunity comparable at least to that of, say, app servers. Unless, of course, the whole thing is just open-sourced.

I've written up most of that argument in http://www.texttechnologies.com/2005/12/11/the-tex t-technologies-market-3-heres-whats-missing/ and related articles. The big missing piece (about the open source alternative) is coming soon, on the same site, and perhaps in my next Computerworld column as well.

One thing in Google's favor, however, is their internal use of knowledge extraction. They seem to really be ahead of the competition in that regard.

Meanwhile, I've got to say -- search is one of the areas where Microsoft has been saber-rattling for a long time, to little effect. Just a couple of quick examples of what I mean:

1. In 1997, I was at the Verity user's conference, and a Microsoft guy there told me how Microsoft would soon be in the business.

2. A few years ago, a woman emailed me and told me she'd just joined Microsoft, and was personally writing all of their web search algorithms.