Yes, creating ontologies is indeed an everyday practise – David Weinberger calls this “lumping and splitting” in his book “Everything is miscellaneous”.

And yes, you’re also right that ISNA has benefitted greatly from the web, as it allowed people from different places to gather around a common concern. Maybe I should be a little bit more optimistic indeed

To take the last thing first, it wasn’t the Sapir–Whorf hypothesis, but the Wikipedia link put me onto the track, it was George Lakoff’s book “Women, Fire, and Dangerous Things”. I don’t think I should enter more deeply into this discussion, is you clearly know a whole lot more abou this than I do.

I think that we build ontologies in our minds every day, like yesterday, when my fiancee told me to buy “the usual diary products”, meaning milk with added Vitamin D and some melon yoghurt.

I don’t think it takes that much to realise that “the usual diary products”, is a class, which is a union of the classes “milk with added vitamin D”, and “melon yoghurt”, which again are subclasses of “milk” and “yoghurt”, respectively. Then, I went out to buy instances of those classes.

A couple of decades from now, thinking this way could well be taught in schools, alongside natural language, after all, I think this is far easier, at this level, than natural language grammar.

Most of the complexity that we, these days, have to struggle with, is to encode this understanding to something that is useful. We lack the tools, we bump into discussions of OWL Full vs. OWL DL, or just SKOS. We do a fair amount of scrambling in the dark, since the complexities are still there in the open. I think that once we’ve scrambled for a few years, the subset of our knowledge and experience that most users will need will emerge, thus, they will be blissfully unaware of all the concerns we’ve had.

So, while I think that, you are right that at the moment, it may be unlikely that groups like ISNA will be able to construct an ontology, it is quite likely that they will in the future. Moreover, and that’s actually a really good example of why I think the Semantic Web has some strong points, is that if they do, they can actually provide an ontology that people can use instead of the false “male/female” binary categorisation that is now everywhere. They can therefore make in impact it is unlikely that they could make by just talking about it.

I might be a little over-optimistic, which may be due to that the education level of my queer friends is higher than the average of all my friends, but I see your point, that historically, marginalised groups has had poorer access to the required education. This is indeed a deep social problem, which must be addressed on several levels, but I do think that the Semantic Web can help rather than deepen the gap, like the web has given even very small groups a voice. Like ISNA, they probably benefitted in some ways from the Web, right?

I admit I was more concerned about the social mediation of technology where we are still dealing with the fact that there are, for instance, very few women in IT, or, to name a more obvious example of silenced voiced, very few queer rights and transgender people, to name just a few “typically marginalized” groups. At the moment it is very unlikely that groups like, for instance, the Intersex society ( http://www.isna.org ) are going to come up with their own ontology, or will be able to provide criticism of already formalized knowledge.

It’s the well-known connex between technological competence and marginalization that is a problem here, and which will have as a result that the (otherwise extremely attractive) rationale of the Semantic Web:

“As long as and ontology is useful to you, then do use it. If it doesn’t reflect your world view, construct one that does.”

won’t work in many case, as the groups who have typically and/or historically been marginalized are typically the ones who lack the skills and resources (or lack of access to people with skills and resources) to create their own ontologies.

This is, of course, not a problem of the Semantic Web approach (and quite on the contrary) but a problem of the social mediation of (any) technology.

In the end, it’s a societal conundrum we haven’t been able to solve yet. While there have been many efforts to, for instance, promote technological awareness among women/girls, we still find that the top three in the list of dream jobs among 15 year old girls (at least in Austria) are: hairdresser, shop assistant, and secretary.

And not even I, even though considered a “geek” among fellow women and/or humanities people, am not (yet!) able to construct my own ontology:-)

So I think that my point doesn’t contradict you, I just meant to point out that there are social issues (and the same old ones mankind has been battling throughout the 20th century) that might pose an obstacle to the concerns of the Semantic Web itself.

Regarding the study your friend brought up: It’s probably got all to do with the Sapir-Whorf hypothesis:

“In linguistics, the Sapir–Whorf hypothesis (SWH) (also known as the “linguistic relativity hypothesis”) postulates a systematic relationship between the grammatical categories of the language a person speaks and how that person both understands the world and behaves in it. Although known as the Sapir–Whorf hypothesis, it was an underlying axiom of linguist and anthropologist Edward Sapir and his colleague and student Benjamin Whorf.” http://en.wikipedia.org/wiki/Sapir-Whorf_hypothesis

Yet I don’t see why this would have the consequence to call the rationale for constructing ontologies into question? For one, an ontology cannot be equated with ‘a model of all the knowledge a language can express#, and secondly, languages are never permanently fixated, as the process of semiosis ( “Semiosis is any form of activity, conduct, or process that involves signs, including the production of meaning. Briefly – semiosis is sign process.” http://en.wikipedia.org/wiki/Semiosis) is never ending.

So the mere confrontation between the Misters Sapir and Whorf and the indigenous people was itself a process that changed and affected the indigenous sign processes they were observing.

Or maybe you can explain?

However, I do not feel this is a show-stopper for the Semantic Web, quite the contrary, it just means that the Semantic Web’s idea that there should not be a single monolithic ontology is the best approach. For one thing, I think the general consensus now is that ontologies for a specific domain is best constructed by the communities that will use them. For example, the general astronomical community should and does construct an ontology for astronomical objects, but it should stay clear of a classification of Active Galactic Nuclei. Since the AGN zoo is pretty complex, that should be left to the more specific AGN community.

I think this also goes for the upper ontologies. There is nothing stopping communities that aren’t like me from creating their own upper ontology, replacing Cyc or UMBEL. Except, of course, that it is a complex undertaking, which may be prohibitively expensive for marginal groups, but I wonder if tool support could solve that problem.

In their applications, marginal groups would then use the ontologies that best reflect their world view. No ontology will be perfect of course, but it may be useful.

There is another point to bring up. A few years ago, a friend of mine brought a study by a famous US linguist of a indigenous group that had only very few “classes” in their vocabulary. My friend felt this very interesting study called into question the very rationale for constructing ontologies. Unfortunately, I do not remember the details, and I’m a programmer, not a linguist…

I feel it is all about usefulness. As long as and ontology is useful to you, then do use it. If it doesn’t reflect your world view, construct one that does. Or if it doesn’t make sense to construct one, because it won’t solve your problem, then don’t.

It is not a show-stopper for the Semantic Web, as the Semantic Web can do a lot of things without relying on ontologies, or even Knowledge Management. For example, the case of “find me the airports with direct flights from me, where it is forecasted good weather the next week”, would not require any such things, it is a case of linking the data points.