On that “shared” conceptualization and other definitions of an ontology

It’s a topic that never failed to generate a discussion on all 10 instalments of the ontology engineering course I taught from BSc(hons) up to participants studying toward or already having a PhD: those pesky definitions of what an ontology is. To top it off, like I didn’t know, I also got a snarky reviewer’s comment about it on my Stuff ontology paper [1]:

A comment that might be superficial but I cannot help: since an ontology is usually (in Borst’s terms) assumed to be a ‘shared’ conceptualization, I find a little surprising for such a complex model to have been designed by a sole author. While I acknowledge the huge amount of literature carefully analyzed, it still seems that the concrete modeling decisions eventually relied on the background of a single ontologist

Is that bad? Does that make the Stuff Ontology a ‘nontology’? And, by the by, what about all those loner philosophers who write single-author papers on ontology; should that whole field be discarded because most of the ontology insights were “shared” only from paper submission and publication?

Anyway, let’s start from the beginning. There’s the much-criticized definition of an ontology from Gruber that, it seems, only novices seem to keep quoting (to my irritation, indeed):

An ontology is a specification of a conceptualization. [2]

If you wonder why quite a bit has been written about it: try to answer what “specification” really means and how it is specified, and what exactly a “conceptualization” is. The real fun starts with Borst et al.’s [3] and then Studer et al.’s [4] refinement of Gruber’s version, which the reviewer quoted above alluded to:

An ontology is a formal, explicit specification of a shared conceptualization. [4]

At least there’s the “formal” (be it in the sense of logic or formal ontology), and “explicit”, so something is being made explicit and precise. But “shared”? Shared with whom? How? Is a logical theory that not one, but two, people write down an ontology, then? Or one person develops an ontology and then emails it to a few colleagues or puts it online in, say, the open BioPortal ontology repository. Does that count as “shared” then? Or is it only “shared” if at least one other person agrees with it as is (all reviewers of the Stuff Ontology did, btw), or perhaps (most or all of) the ‘conceptualization’ of it but a few axioms would need a bit of tweaking and cleaning up? Do you need at least a group of people to develop an ontology, and if so, how large should that group be, and should that group consist of independent sub-groups that adopt the ontology (and if so, how many endorsers)? Is a lightweight low-hanging-fruit ontology that is used by a large company a real or successful ontology, but a highly axiomatised ontology with a high tangledness that is used by a specialist organization, not? And even if you canvass and get a large group and/or organization to buy into that formal explicit specification, what if they are all wrong on the reality is supposed to represent? Does it still count as an ontology no matter how wrong the conceptualization is, just because it’s formal, explicit, and shared? Is a tailor-made module of, say, the DOLCE ontology not also an ontology, even if the module was made by one person and made available in an online repository like ROMULUS?

Perhaps one shouldn’t start top-down, but bottom-up: take some things and decide (who?) whether it is an ontology or not. Case one: the taxonomy of part-whole relations is a mini-ontology, and although at the start only ‘shared’ with my co-author and published in the Applied Ontology journal [5], it has been used by quite a few researchers for various (and unintended) purposes afterward, notably in NLP (e.g., [6]). An ontology? If so, since when? Case two: Noy et al. converted the representation of the NCI thesaurus into OWL DL [7]. Does changing the serialisation of a multi-authored thesaurus from one format into another make it an ontology? (more on that below.) Case three: a group of 5 people try to represent the subject domain of, say, breast cancer, but it is replete with mistakes both regarding the reality it ought to represent and unintended modelling errors (such as confusing is-a with part-of). Is it still an ontology, albeit a bad one?

It gets more muddled when the representation language is thrown in (as with case 2 above). What if the ontology turns out to be unsatisfiable? From a logic viewpoint, it’s not a theory then (a consistent set of sentences, is), but if it’s formal, explicit, and shared, is it acceptable that those people who developed the artefact simply have an inconsistent conceptualization and that it still counts as an ontology?

Horrocks et al. [8] simplify the whole thing by eliminating the ‘shared’ aspect:

an ontology being equivalent to a Description Logic knowledge base. [8]

However, this generates a set of questions and problems of its own that are practically also problematic. For instance: 1) whether transforming a UML Class Diagram into OWL ‘magically’ makes it an ontology (answer: no); 2) The NCI Thesaurus to OWL (answer: no); or 3) if you used, say, Common Logic to represent it, that then it could not be an ontology because it’s not formalised in Description Logics (answer: it sure can be one).

There are more attempts to give a definition or a description, notably by Nicola Guarino in [9] (a key paper in the field):

An ontology is a logical theory accounting for the intended meaning of a formal vocabulary, i.e. its ontological commitment to a particular conceptualization of the world. The intended models of a logical language using such a vocabulary are constrained by its ontological commitment. An ontology indirectly reflects this commitment (and the underlying conceptualization) by approximating these intended models. [9]

That’s a mouthful, but at least no “shared” in there, either. And, finally, among the many definitions in [10], here’s Barry Smith and cs.’s take on it:

An ONTOLOGY is a representational artifact, comprising a taxonomy as proper part, whose representational units are intended to designate some combination of universals, defined classes, and certain relations between them. [10]

And again, no “shared” either in this definition. Of course, also with Smith’s definition, there are things one can debate about and pose it against Guarino’s definition, like the “universals” vs. “conceptualization” etc., but that’s a story for another time.

So, to sum up: there is that problem on how to interpret “shared”, which is untenable, and one just as well can pick a definition of an ontology from a widely cited paper that doesn’t include that in the definition.

That said, all this doesn’t help my students to grapple with the notion of ‘an ontology’. Examples help, and it would be good if someone, or, say, the International Association for Ontology and its Applications (IAOA) would have a list of “exemplar ontologies” sooner rather than later. (Yes, I have a list, but it still needs to be annotated better). Another aspect that helps explaining it comes is from Guarino’s slides on going “from logical to ontological level” and on good and bad ontologies. This first screenshot (taken from my slides—easier to find) shows there’s “something more” to an ontology than just the logic, with a hint to reasons why (note to my students: more about that later in the course). The second screenshot shows that, yes, we can have the good, bad, and ugly: the yellow oval denotes the intended models (what it should be), and the other ovals denote the various approximations that one may have tried to represent in an ontology. For instance, representing ‘each human has exactly one brain’ is more precise (“good”) than stating ‘each human has at least one brain’ (“less good”) or not saying anything at all about it an ontology of human anatomy (“bad”), and even “worse” it would be if that ontology ware to state ‘each human has exactly two tails’.

Maybe we can’t do better than ‘intuition’ or ‘very wieldy explanation’. If this were a local installation of WordPress, I’d have added a poll on definitions and the subjectivity on the shared-ness factor (though knowing well that science isn’t governed as a democracy). In lieu of that: comments, preferences for one definition or the other, or any better suggestions for definitions are most welcome! (The next instalment of my Ontology Engineering course will start in a few week’s time.)

 

References

[1] Keet, C.M. A core ontology of macroscopic stuff. 19th International Conference on Knowledge Engineering and Knowledge Management (EKAW’14). K. Janowicz et al. (Eds.). 24-28 Nov, 2014, Linkoping, Sweden. Springer LNAI vol. 8876, 209-224.

[2] Gruber, T. R. A translation approach to portable ontology specifications. Knowledge Acquisition, 1993, 5(2):199-220.

[3] Borst, W.N., Akkermans, J.M. Engineering Ontologies. International Journal of Human-Computer Studies, 1997, 46(2-3):365-406.

[4] Studer, R., Benjamins, R., and Fensel, D. Knowledge engineering: Principles and methods. Data & Knowledge Engineering, 1998, 25(1-2):161-198.

[5] Keet, C.M., Artale, A. Representing and Reasoning over a Taxonomy of Part-Whole Relations. Applied Ontology, 2008, 3(1-2):91-110.

[6] Tandon, N., Hariman, C., Urbani, J., Rohrbach, A., Rohrbach, M., Weikum, G.: Commonsense in parts: Mining part-whole relations from the web and image tags. In: Proceedings of the Thirtieth AAAI Conference on Artificial Intelligence (AAAI’16). pp. 243-250. AAAI Press (2016)

[7] Noy, N.F., de Coronado, S., Solbrig, H., Fragoso, G., Hartel, F.W., Musen, M. Representing the NCI Thesaurus in OWL DL: Modeling tools help modeling languages. Applied Ontology, 2008, 3(3):173-190.

[8] Horrocks, I., Patel-Schneider, P. F., and van Harmelen, F. From SHIQ and RDF to OWL: The making of a web ontology language. Journal of Web Semantics, 2003, 1(1):7.

[9] Guarino, N. (1998). Formal ontology and information systems. In Guarino, N., editor, Proceedings of Formal Ontology in Information Systems (FOIS’98), Frontiers in Artificial intelligence and Applications, pages 3-15. Amsterdam: IOS Press.

[10] Smith, B., Kusnierczyk, W., Schober, D., Ceusters, W. Towards a Reference Terminology for Ontology Research and Development in the Biomedical Domain. KR-MED 2006 “Biomedical Ontology in Action”. November 8, 2006, Baltimore, Maryland, USA.

Book reviews for 2016

I can’t resist adding another instalment of brief reviews of some of the books I’ve read over the past year, following the previous five editions and the gender analysis of them (with POC/non-POC added on request at the end). This time, there are three (well, four) non-fiction books and four fiction novels discussed in the remainder of the post. The links to the books used to be mostly to Kalahari.com online (an SA-owned bookstore), but they have been usurped by the awfully-sounding TakeALot, so the links to the books are diversified a bit more now.

Non-fiction

Writing what we like—a new generation speaks, edited by Yolisa Qunta (2016). This is a collection of short essays about how society is perceived by young adults in South Africa. I think this stock-taking of events and opinions thereof is a must-read for anyone wanting to know what goes on and willing to look a bit beyond the #FeesMustFall sound bites on Twitter and Facebook. For instance, “A story of privilege” by Shaka Sisulu describing his experiences coming to study at UCT, and Sophokuhle Mathe in “White supremacy vs transformation” on UCT’s new admissions policy, the need for transformation, and going to hold the university to account; Yolisa Qunta’s “Spider’s web” on the ghost of apartheid with the every-day racist incidents and the anger that comes with it; “Cape Town’s pretend partnership” by Ilham Rawoot on his observations of exclusion of most Capetonians regarding preparations of the World Design Capital in 2014. There are a few ‘lighter’ essays as well, like the fun side of taking the taxi (minibus) in “life lessons learnt from taking the taxi” by Qunta (indeed, travelling by taxi can be fun).

Elephants on Acid by Alex Boese (2007). This is a fun book about the weird and outright should-not-have-been-done research—and why we have ethics committees now. There are of course the ‘usual suspects’ (gorillas in our midst, Milgram’s experiment), the weird ones (testing LSD on elephants; didn’t turn out alright), funny ones (will your dog get help if you are in trouble [no]; how much pubic hair you lose during intercourse [not enough for the CSI people]; social facilitation with cockroach games; trying to weigh the mass of a soul), but also those of the do-not-repeat variety. The latter include trying to figure out whether a person under the guillotine will realise it has been ‘separated’ from his body, Little Albert, and the “depatterning” of ‘beneficial brainwashing’ (it wasn’t beneficial at all). The book is written in an entertaining way, either alike a ‘what on earth was their hypothesis to devise such an experiment?’, or, knowing the hypothesis, with some morbid fascination to see whether it was falsified. Most of the research referenced is, for obvious reasons, older. But well, that doesn’t mean there wouldn’t be any outrageous experiments being conducted nowadays when we look back in, say, 20 years time.

What if? by Randall Munroe (2014, Dutch translation, dwarsligger). Great; read it. Weird and outright absurd questions asked by xkcd readers are answered sort of seriously from a STEM perspective.

Say again? The other side of South African English by Jean Branford and Malcolm Venter (2016). This short review ended up a lot longer, so it got its own blog post two weeks ago.

Fiction

Red ink by Angela Makholwa (2007). This is a juicy crime novel, as the Black Widow Society by the same author is (that I reviewed last year), and definitely a recommendable read. The protagonist, Lucy Khambule, is a PR consultant setting up her company in Johannesburg, but used to be a gutsy journalist who had sent a convicted serial killer a letter asking for an interview. Five years hence, he invites her for that interview and asks her to write a book about him. As writing a book was her dream, she takes up the offer. Things get messy, partly as a result of that: more murders, intrigues, and some love and friendship (the latter with other people, not the serial killer) that put the people close to Lucy in harm’s way. As with the Black Widow Society, it ends well for some but not for others.

Things fall apart by Chinua Achebe (1958 [2008 edition]). This is a well-known book in Africa at least, and there are many analyses are available online, so I’m not going to repeat all that. The story documents both the mores in a rural village and how things—more precisely: the society—fall apart due to several reasons, both on how the society was organised and the influence of the colonialists and their religion. The storytelling has a slow start, but picks up in pace after a short while, and it is worthwhile to bite through that slow start. You can’t feel but a powerless onlooker to how the events unfold and sorry how things turn out.

Kassandra by Christa Wolf (1983, Dutch translation [1990] from the German original; also available in English). Greeks, Trojans, Achilles, Trojan Horse, and all that. Kassandra the seer and daughter of king Priamos and queen Hadebe, is an independent woman, who rambles on analysing her life’s main moments before her execution. It has an awkward prose that one needs to get used to, but there are some interesting nuggets. On only approaching things in duals, or alternative options, like endlessly win or loose wars or the third option of to live. It was a present from the last century that I ought to have read earlier; but better late than never.

De midlife club by Karin Belt (2014, in Dutch, dwarsligger). The story describes four women in their early 40s living in a province in the Netherlands (the author is from a city nearby where I grew up), for whom life didn’t quite turn out as they fantasised about in their early twenties, due to one life choice after another. Superficially, things seem ok, but something is simmering underneath, which comes to the surface when they go to a holiday house in France for a short retreat. (I’m not going to include spoilers). It was nice to read a Dutch novel with recognisable scenes and that contemplates choices. The suspense and twists were fun such that I really had to finish reading it as soon as possible.

As I still have some 150 pages to go to finish the 700-page tome of Indaba, my children by Credo Mutwa, a review will have to wait until next year. But I can already highly recommend it.