Talk:Ontology (information science)

From Wikipedia, the free encyclopedia

	This article is within the scope of WikiProject Computer science, which aims to create a comprehensive computer science reference for Wikipedia. Visit the project page for more information and to join in on related discussions.
???	This article has not yet received a rating on the assessment scale.
???	This article has not yet received an importance rating on the assessment scale.

[edit] terminology

An ontology is not a 'data model', as specified in the first line of the entry. Data models necessitate formal mechanism's for providing structure, integrity and manipulation. An ontology as a conceptual entity only enforces structure, and as such its description as a data model is misleading - it is a data structure/schema.

I agree. It is not a data model but a model for knowledge representation. —Preceding unsigned comment added by 212.114.205.92 (talk) 12:06, 12 November 2007 (UTC)

Just a question on terminology: why not use 'generic ontology' instead of foundation ontology or upper ontology? MajFreak

I've seen Top Level Ontology too?

--130.76.32.167 (talk) 02:43, 12 February 2008 (UTC) I also agree. Some ontologies may be used as data models, but some are not. So this is misleading and inaccurate. It would be correct to say an ontology is a model. So this can be fixed by merely removing the word data in the definition. [Dr. Michael Uschold]

[edit] Entities/concepts

Q: Should this line: "The purpose of a computational ontology is not to specify what does or does not 'exist', but to create a database, which is a human artifact, containing concepts referring to entities of interest to the ontologist (...)" read "(...) containing entities referring to concepts of interest (...)", or am I wrong? I thought databases contained entities, not concepts?

A: DBs contain elements (entities, objects, ...): if you're describing an ontology, expecially a "core ontology", you're referring to concepts, categories, not necessary to particular objects. In short, it depends on how you interpret them (and what they refer to). In an RDF statement, for example, you talk about "resources", which can be URIs, texts, even other statements. Bye, Alessandro

Good introductory article! I just regret that the editor(s) who quoted authors like T. R. Gruber and Peter Murray-Rust did not actually cite the appropriate references! --Alexandre 21:38, 2 Jul 2004 (UTC)

I changed this to saying it was a conceptualisation not a database, correct me if i'm wrong, but the database that we form ontologies in are purely a side effect of having to store the conceptualisation --RickiRich 00:17, 20 November 2005 (UTC)

[edit] Article refactoring

In my point of view, the article stresses the upper level ontology way to much and misses a lot of other topics. I am working at a research institute (mainly dealing with ontologies) and could probably (with the help of my colleagues) rewrite the article quite a bit. What would be a good process to do so? Just do it? Post it on a web-page? How do you "suggest" articles? As I would really like to remove quite a lot from the articel, I fear I could offen the original authors or the community? As I am new to editing wikipedia pages, I would LOVE to get some advices for my problems. I You have any, You can post them here or email them to mvo at aifb.uni-karlsruhe.de If this is the wrong place to ask for help, feel free to remove my question completely :-) Thank You, Max.

I would suggest that you first post here on the discussion page a proposal for a new outline and ask the other contributors for their opinion. If after let's say three weeks there are no substantial objections then just go ahead and reorganize and expand it. From my point of view the article contains useful information. The areas for further work are the parts on Anatomy of an ontology and Ontology languages? And it would be nice to have two elaborated examples. Hirzel 15:42, 18 Jun 2005 (UTC)

Just do it. If you're concerned that it will be controversial, or if it'll be initially less "polished" than the current article, or if you'd prefer to get initial comments on a rough outline, then alternatively place it at a temporary page. (Say, a sub-page of your user page, or an article-space page Ontology (computer science)/Temp.) Alai 15:48, 18 Jun 2005 (UTC)

Folks, I'd hope we could keep in mind the Wikipedia guidlines for controversial topics and keep arguments for and against upper or common ontology in their respective sections. We can agree to disagree on these things, but respect the opinions of others. If topics are missing, I'd say feel free to add them, but not to delete prose that others have taken the time to add and feel is important. apease

Some thoughts ... Besides the upper-generic vs domain debate, there are other debates, for example between the "pragmatic" view (ontologies as engineering artefacts and social agreements), represented e.g. by Tom Gruber, and the "scientific" view (ontologies as scientific representation of the real world), represented e.g. by the European Centre for Ontological Research. Other debates are whether the subsumption relationship is fundamental, or the partitive one, etc. In that highly controversial context on what ontology is or ontologies are (using singular or plural is also part of the debate), it's clear that e.g. the assertion "an ontology is the product of an attempt to formulate an exhaustive and rigorous conceptual schema about a domain" is right away choosing camp in the "pragmatic" vs "scientific" debate. So, seems to me that the article should start by stressing that there is currently no consensus on what an ontology is or should be, then present the different viewpoints with their pro-con arguments and links to representative resources. Also, what is lacking is what ontologies are made for, since they are basically tools with purposes. Maybe the article should start by this aspect. universimmedia

[edit] Italian translation

I have added the link to the italian translation for this page. I am planning to add some content to the italian page, then I will try to back-translate this stuff in english.

I hope the grammar will be acceptable!

--Mcoletti 14:47, 15 Sep 2004 (UTC)

[edit] Wiki categories

Could you mention how much (or few) ontologies are related to Wikipedia categories? -- Error 01:03, 22 Nov 2004 (UTC)

The article includes a particular view against the viability of an upper ontology. There are quite a few proponents of an upper ontology. Does anyone object to the addition of a reasoned counterpoint to the view currently expressed? User:Apease

[edit] Language and Concepts

"It is essential to separate the language used to refer to terms from the terms themselves." ??? This sentence either needs some critical clarification or it needs to be removed. This also points to a fundamentally mistaken assumption at work in the section titled "Why an Upper Ontology is Feasible." Namely, that an ontology can provide a mapping of concepts that is faithful to a common understanding of these concepts without reintroducing the ambiguous, culturally relativistic, self-reflexive, imprecise characteristics that are inherent in language. Last time I checked, tools like SUMO are still using words (mostly English) to represent concepts. So I don’t think it is true that the “conflation of ontology, language and knowledge” is an error that can be avoided. This conflation is simply an inevitable fact of human existence. What rational justification is there for believing that concepts can exist outside of their linguistic formulations anyway?? The conceptual landscape is determined and constrained by language. There are a lot of things we can do with language. But one thing we can’t do is stand outside of all linguistic formulations and talk about concepts in general.

REPLY: SUMO does not use words to represent concepts. It uses words as convenient labels for concepts. One rational justification for this is the existence of translations of SUMO concepts into multiple non-English languages. Reasonable people simply disagree about this issue.

KL: So, I'm still not clear how using "words as convenient labels for concepts" is different from "using [English] words to represent concepts." The implication here is that SUMO embodies a completely transparent representation of concepts which entirely circumvents the messy reality that there is no way to describe, define, characterize, organize, etc., concepts without employing human language. Chinese and Hindi versions of SUMO are no evidence that this transparent representation has been achieved. Rather, there are now two more representative approximations of a particular conceptual schema using radically different symbolic/linguistic idioms. Creators and users of the Chinese SUMO may have reached some kind of consensus, but that’s no basis for claiming that there is scientific precision in the whole set of “term A = concept B” formulations.

Leaving aside the inexact science of translation, however, the lynchpin criticism of SUMO here is that it can never embody anything more than one conceptual schema, which presupposes one epistemic attitude and/or one theoretical framework for the whole structure and contents of the upper ontology. Plato tried something similar with his realm of ideal ‘forms.’ The forms are envisioned as absolute unchanging concepts of pristine purity, unsullied by the particulars of language or real-world contexts. And because the ideal form of ‘dog’ contains only the essential qualities of “dogness,” the ideal form becomes a template enabling our recognition of all of the real-world instances of “dogs.” There is a long and fairly damning history (2000+ years) of criticism of the theory of the Forms, starting with Aristotle and his “third man” argument.

Following Richard Nisbett and his book The Geography of Thought, I have a hard time conceiving of a single ontology for both Western and Asian thinkers. The ancient Greek passion for abstract categories into which the entire world can be taxonomically arranged is prototypically Western. The whole notion of causality is essentially Western. “In the Chinese intellectual tradition there is no necessary incompatibility between the belief that A is the case and the belief that not-A is the case. On the contrary, in the spirit of the Tao or yin-yang principle, A can actually imply that not-A is also the case, or at any rate soon will be the case…. Events do not occur in isolation from other events, but are always embedded in a meaningful whole in which the elements are constantly changing and rearranging themselves. [In the Chinese approach to reasoning,] to think about an object or event in isolation and apply abstract rules to it is to invite extreme and mistaken conclusions. It is the Middle Way that is the goal of reasoning.”

A good exercise for any formal system is to replace the symbols with arbitrary strings. If you get the same inferential closure from the system without resorting to the implie meaning of the labels, then the full semantics of the system has been defined and does not depend on language. That's the case for SUMO just as it is for arithmetic. The meaning of "plus" or "+" doesn't depend on language but rather the formal semantics of the symbol.

[edit] Ontology vs. taxonomy

i added the expand tag because the article mentions that ontologies have certain concepts that make them different from simple taxonomies, but it does not yet do a good job in explaining what theses differences are. 790 08:52, 11 Apr 2005 (UTC)

Correct. See talk:folksonomy for more issues of this kind.

Section "Domain ontologies and upper ontologies" says that WordNet is not an ontology, but section "Relation to the philosophical term" lists WordNet as an example for an ontology. 195.212.29.179 (talk) 09:00, 29 April 2008 (UTC)

[edit] Order of things

Maybe better if the more regular form of ontologies (domain ontology) is explained in-depth before the upper ontologies.

Agreed. Also the fact that a strong ontology is distinguished from a weak ontology based on the ability to infer things from a strong but not a weak one: a weak one requires human judgement to draw inferences, and this can't be done automatically.

[edit] This intro is garbage.

In information science, an ontology is the product of an attempt to formulate an exhaustive and rigorous conceptual schema about a domain. An ontology is typically a hierarchical data structure containing all the relevant entities and their relationships and rules within that domain (e.g., a domain ontology). However, computational ontology, like any ontology, does not have to be hierarchal at all.

it seems fine to here, and becomes garbage here:

Due to the cyclical nature of the universe itself, ontology is better served with cycles. Hierarchy is not present in spatial contexts, so as Aristotle would say (parroting his master), microcosm as macrocosm; computational reliance on hierarchy is perhaps a stepping stone to greater things.

then it cleans itself up:

The computer science usage of the term ontology is derived from the much older

usage of the term ontology in philosophy.

An ontology which is not tied to a particular problem domain but attempts to describe general entities is known as a foundation ontology or upper ontology. Typically, more specialized domain specific schemata must be created to make the data useful for real world decisions.

Okay, I still don't have any idea what a CS Ontology is. Someone needs to put away the thesaurus and write in plain English. Talking about the "Universe's cyclical nature" is garbage. Stating there is no hierarchy in spatial contexts is obtuse. Talking about aristotle and cute little "parroting his master" asides is garbage. This whole thing needs to be thrown out and rewritten by someone whose goal is to teach instead of some pathetic attempt to show off.

All agreed, but the only possible way to introduce it is with examples that make clear when a weak ontology is inadequate and stronger ones evolve as the weak ones prove deficient. That discussion is mostly at folksonomy and talk:folksonomy and parallels that on folk taxonomy and taxonomy.

Fact is, the computer science geeks stole this term from philosophers and they abuse it badly. Sorting out the mess requires making distinctions like strong (programs do the inferences) versus weak (users do the inferences).

[edit] thoughts

The first sentance is untrue, the second sentence contradicts the third, and I think uses "ontology" in two different ways without explaining.

"Fact is, the computer science geeks stole this term from philosophers and they abuse it badly. Sorting out the mess requires making distinctions like strong (programs do the inferences) versus weak (users do the inferences)."

Irrespective of how it has been 'abused' with relation to philosphy, it is a useful and formally described concept in computer science. I too, do not like the way it is explained here. To give pre-warning, I will probably re-write at least the intro, and move the parts on Upper Ontologies to their own page. Currently 'Uses for Ontologies' doesn't mention the Semantic web, I'll add that. Comments?

If you'd like to make a section on how CS Ontologies are abused forms of Philosophy Ontologies, I'd be interested in reading it. My view is that however they may be an abused form of the Philosophy ontology, CS ontologies are useful and practical.

--RickiRich 03:24, 19 November 2005 (UTC)

[edit] I edited it

I took out 'cyclic nature' and the information on Aristotle, If it wants to be here (which if your explaining a computer science term, I don't think it does), its in the vs Philosophy section). I edited the introduction, and extended explanation that the "rigorous and exhaustive" is only in a certain domain of interest. I filled out Ontology (computer_science)#Strong vs. weak ontologies. I changed part of

"Ontology in computing vs. ontology in philosophy

This is different from - but related to - the philosophical meaning of the word ontology, the study of existence. The purpose of a computational ontology is not to specify what does or does not 'exist', but to create a conceptualisation of the concepts and relationships in the area of interest to the ontolgist. The reasoning used by philosophical ontologists can be helpful in recognizing and avoiding potential logical ambiguities."

from "but to create a database" to "but to create a conceptualisation" because the purpose is to create a conceptualisation, not a database (a database is a side effect).

I moved the Upper Ontology disussions to Upper Ontologies (computer science).
I moved the example Upper Ontologies (computer science) to that page.
I added "Upper Ontologies have uses such as allowing the a computer to infer information from a larger domain of knowledge from usual." to "Uses for ontologies". It may be useful but its not written well, please change if you've got more skills.
I added Upper Ontologies (computer science) and Foundation Ontology to the See Also list.
--RickiRich 00:09, 20 November 2005 (UTC)

[edit] Merging Upper Ontology with this page

Hi R.Koot, why is it you think the two pages should be merged?
I'll happily be shown differently, but it seemed to me that Upper Ontologies took up a large portion of the Ontology page, and also was slightly confusing. As far as I am aware, upper ontologies are not always used when ontologies are used (Tim Berners-Lee doesn't necessarily want to make 'one big ontology'). Those were the reasons I put it on a different page :)

Keep separate There is plenty of material to justify separate articles. RayGates 01:23, 6 March 2006 (UTC)

[edit] Types

In the first section at the very end there it says: <<Thus the need for standards which take 'core' ontologies (e.g. the Dublin Core in SGML) and solidify them into 'foundations'.>> Shouldn't it be <<Thus the need for standards (e.g. the Dublin Core in SGML) which take 'core' ontologies and solidify them into 'foundations'.>> Normen.mueller 10:38, 6 March 2006 (UTC)

[edit] Edit to intro

Why I made this edit to the intro:

An ontology is not necessarily about the "real world" -- it can be about a fictional world, etc.
An ontology is not necessarily about a "specific part" of a world. It can aim to represent every aspect of a world (at some level of abstraction, of course).

dbtfz ^talk 07:05, 16 March 2006 (UTC)

I certainly agree with your first point, that was an oversight on my part. However I'm not certain of the second. While this is true in theory (perhaps we could cite Cyc as an attempt to do this for the "real world"), in practice I argue that all ontologies are only partial. Even your point about ontologies being abstractions of a world makes this case: abstraction is a process that reduces the scope of the knowledge in an ontology.

Defining "ontology" is hard because it is easy to be so general that the definitions is drained of all useful meaning. How do we define "ontology" in a way that is different from say a simple data model? Problems, problems :-) Gwernol 16:04, 16 March 2006 (UTC)

Thanks for the comment. I disagree with your point about my second point, though I think the disagreement may be merely verbal. The way I use the term "represent," a street map of (say) New York City represents all of the city, even though it does not capture every aspect of the city (e.g. the height of the buildings). Similarly an ontology like (say) the Cyc ontology represents (or at least aims to represent) the entire world, even though it does not capture every aspect of it. You may be using "part" to mean what I mean by "aspect". If what you're saying is that all ontologies are partial in that they cannot capture every aspect of the domain they represent, I fully agree. I just don't think that an ontology has to capture every aspect or detail of a domain in order to be said to represent that domain. Anyway, you should certainly feel free to edit the intro as you see fit. A little back-and-forth on this can only improve the article. dbtfz ^talk 03:38, 17 March 2006 (UTC)

I have radically simplified the introduction. It had drifted off into a strange philosophical discourse that was obfuscating rather than illuminating. It also had a long list of alternative meanings for "ontology" that added nothing to the reader's understanding of what an ontology is. I'm sure there's room for further improvements, but I think its much better (especially in an introduction) to have a clear, concise description of what an ontology is.

I'd love to hear what others think. Gwernol 15:36, 24 March 2006 (UTC)

It looks great to me. Very clear and concise. Nice work! dbtfz ^talk 01:33, 25 March 2006 (UTC)

I agree, its concise and informative --RickiRich 15:33, 25 March 2006 (UTC)

[edit] Further improvements to the article

Thanks for your kind remarks about my edits to the intro section. Emboldened by your comments, I have started editing the rest of the article. I'm trying to make it clearer for non-expert readers, assuming that ontologists and other experts in the field already understand the concepts being explained. In particular I'm trying to use examples to illustrate the various points as I go. As always, these are just my thoughts on how to do this. Gwernol 18:32, 25 March 2006 (UTC)

I'm also being bold and making a few changes. If any of these are controversial, feel free to revert them, revise them, and/or start a discussion here. Speaking of "bold," I think the article overuses boldface text, and I'm going to remove some of this. dbtfz ^talk 03:59, 18 April 2006 (UTC)

[edit] Strong vs. weak ontologies

A few comments regarding this subsection:

Can anyone provide a good source supporting the "strong" vs. "weak" distinction? It's a perfectly sensible distinction, but I'm not sure how standard it is or how commonly it is made in these terms.
I'm not a big fan of the claim that "A paragraph of written text could technically be defined as a weak ontology as it contains knowledge about the world." Surely containing knowledge about the world is not sufficient for being an ontology? Wouldn't that make the concept of ontology practically vacuous (as nearly every information artifact contains knowledge about the world)? (On the other hand, it would be quite sensible to claim that a paragraph of written text presupposes or implicitly makes use of an ontology.)
I'm also not too keen on the claim that "An ontology defined in a formal ontology language such as OWL is a strong ontology, as a computer can understand the complete meaning (or semantics) of the ontology." "Understand" seems too strong. It is extremely controversial whether computers can understand anything at all--in principle let alone in practice. Yet what we're saying here ought not to be controversial. It seems like what we want to say is that such an ontology is formulated in a precise way that is relatively easy for a computer to process, or something like that. (There is already a reference to machine-readability, which is on the right track.)

I don't intend any of this to be dogmatic or harshly critical. Just a few things to consider. dbtfz ^talk 01:36, 19 April 2006 (UTC)

Dbtfz, I agree with you. I kept this section when I revised the article because it does seem like a useful distinction. However after reading your observations I am swayed by your arguments, especially about natural language not being an ontology in any meaningful sense. I would support removing the whole strong vs. weak distinction from the article unless someone else can find good sources that show it is in common use in the field. I also agree that the sentence you highlight about formal ontology languages is too strongly worded. You should go ahead and re-write it. Good work, Gwernol 01:44, 19 April 2006 (UTC)

Thanks, Gwernol. I went ahead and removed the "strong vs. weak" subsection, but added some content to the intro which I think expresses the basic idea in a clearer and more appropriate way. (Really, a CS ontology must be machine-readable, i.e. "strong," if it is to be of any use.) I went out on a limb in saying that there are two main distinctions between philosophical ontologies and computer science ontologies (I don't have a source for that claim; it just seems right and fairly obvious to me), but that's open to revision of course. If there are any other crucial distinctions, they should be added. dbtfz ^talk 02:09, 19 April 2006 (UTC)

[edit] More changes April 2006

I've rearranged the page and removed some redundant/sloppy content that was contributing little to the article. In particular, I moved the "Anatomy of an ontology" section to the top, since it covers the basics of what an ontology is. Regarding this section:

Instead of breaking ontologies down into concepts/attributes/relations (with concepts broken down into concrete/abstract), I would like to break them down into individuals/classes/attributes/relations, as I've already done in the intro.
I don't like the term "concept" as its being used here, because it's very odd to think of, say, my cat or my desk as a concept. "Object" is much better, I think.
I don't agree with the characterization of "concepts" that are individual/non-set-like as being "concrete". There are lots of abstract individual/non-set-like-objects (e.g. the number 1).
It should be noted that in some ontologies, such as Cyc, what we're referring to as "attributes" here are really just special kinds of relations, where one or more of the relata are quantities or other things that are less natural to characterize as full-fledged "objects". Whether this holds for an ontology is dependent, I suppose, on whether that ontology treats quantities (e.g. 5kg, 20 meters) as first class objects. (Cyc, for example, does.)

I have many more thoughts on this, but that's enough for now. :-) dbtfz ^talk 03:00, 19 April 2006 (UTC)

Made some more bold changes today, and plan on making more along these lines in the near future. I hope they will be seen as improvements. If anyone has a different vision for how the article should be structured or presented, I'm open to discussion. dbtfz ^talk 02:18, 21 April 2006 (UTC)

Generally I like the clarifications you made in section 1. I am not yet sure if I like the replacement of concept by object. The term 'object' has connotations which restrict the scope of understanding. People are writing ontologies about concepts they think of and not directly about the objects. I like to emphasize keep the association to concept maps which are basically graphical representations of ontologies. Hirzel 09:15, 21 April 2006 (UTC)

Thanks for the feedback. I understand your concerns, and I think we can reach a compromise. More comments on this soon when I have more time.... dbtfz ^talk 15:36, 21 April 2006 (UTC)

If anyone wants to change the references to "objects" back to "concepts," that's fine with me. I would eventually like to have something in the article explaining and justifying the use of this term, however. One must concede, I think, that it is, prima facie, a category mistake to refer to, say, a specific Ford Explorer as a "concept". dbtfz ^talk 00:10, 24 April 2006 (UTC)

[edit] Inconsistency bug in "Relationships" section

The text and graphical versions of the example are inconsistent, i.e. the Ford Explorer is described as a 4-wheel drive in a text but is shown as a 2-wheel drive in the corresponging figure.

I would change it myself if I knew the answer - however the names "Bronco" and "Explorer" are not used by Ford Europe

Cord Hockemeyer

Thanks Cord, well spotted. The text is correct, the diagram is wrong. I'll correct the diagram momentarily. Gwernol 16:05, 12 May 2006 (UTC)

I've fixed it. You will likely need to refresh your browser to see the updated image. Gwernol 16:14, 12 May 2006 (UTC)

Considering the same diagram:

Shouldn't the number of drive wheels be a superclass of Car? That is, a Car is a 2-Wheel Drive [thing], not the other way around. Following this diagram, either Truck would have it's own subclasses of two- and four-weel-drive, or it would have to share the subclasses of Car, neither of which seems right. —Ben FrantzDale 16:50, 8 June 2006 (UTC)

Ah now that's a very different sort of question, and an interesting one. This is a very common problem in ontology construction and maintenance. Should a particular attribute be expressed as a sub-class (i.e. should "4-wheel drive" be a class of car) or is it a super-class (i.e. "car is a type of 4-wheel vehicle"). There is no correct general answer. You build the ontology depending on the context in which the ontology is being used. For example, if our vehicle ontology was mainly being used by consumers purchasing a car, then you probably want "4-wheel drive" to be a sub-class of car, since that is the main focus of the ontology. For a more general (Cyc-like) ontology you may well be correct.

Be aware too, that generalizing concepts up the tree (in this example, making 4-wheel drive a super-class of both Truck and Car) can have its problems too. Is 4-wheel drive really the same concept for both cars and trucks? In this example it probably is, but there are often subtle but important differences in apparently similar concepts which mean you should be careful of over-generalization.

These are the sorts of questions that professional ontologists worry over all day long and which make developing large-scale ontologies hard. Gwernol 17:28, 8 June 2006 (UTC)

[edit] Minor remark on the last comment

The simple and correct solution is the rephrase those concepts as a "2-wheel drive car", and a "4-wheel drive car", and to remodel-relate accordingly. An ontology is not about exploding a hierarchy with different classification views... A concept termed 2-wheel drive is therefore an empty/nonsense and even obsolete concept, it is a concept that has not been thought over, e.g. that disregards subsumption and some other basic modeling principles. Daviddecraene 12:17, 14 June 2006 (UTC)

I think you misunderstand my point David. I was arguing that whether the concept of "2-wheel drive" etc. was relevant depends on the purpose of that particular ontology you are working on. For some purposes your solution would be the correct one, in other contexts a diffrent solution would be appropriate. There is no canonically correct answer. Gwernol 11:44, 15 June 2006 (UTC)

I'll admit I'm a novice, on the subject, but to me the diagram seems to achieve its end; it illustrates the concept of categorization. If you really must have a different example, why not divide plasticware? First, you have spoons and forks, and divide the spoons into red spoons and white spoons. *shrug* Luna Santin 11:58, 15 June 2006 (UTC)

[edit] On the Distinction between ontology in computer science & ontology in philosophy

The distinction that is maintained in wikipedia between ontology used in computer science and ontology in philosophy is, in my opinion, nonsense. There is no black and white distinction between the two. The only difference that can exist is that of a formalised ontology versus an ontology that is not explicitely based on a formal upper layer or philosophical principles (ontology in it's broadest definition). In my opinion both articles should be integrated into a complete article on ontology. Daviddecraene 11:39, 15 June 2006 (UTC)

Again, I disagree. The philosophical endeavour of ontology is a the study of the categories of existence. In computer science the term ontology is used to mean the specific set of knowledge strutures used to represent (some part of) a domain. Its really a particular database instantiation. This is such a different meaning that its unfortunate that the same term is used for both. Personally I prefer "knowledge base" for the computer science usage, although that has its own synonym problems, unfortunately.

Gwernol 11:48, 15 June 2006 (UTC)

In computer science the term ontology is indeed used very broadly and sometimes used improperly from a philosophical perspective. This term misuse is unfortunate, but to say that ontologies in computer science are never ontologies as philosophy intends is however also incorrect. If this article is not about ontologies applied to computer science, but about knowledgebases, terminologies, a lexicon, a taxonomy or any database then this article should be renamed. However I get the impression that this is indeed ontology applied, as it describes an attempt to classify enduring things (cfr section classes, and in particular, our debated figure on cars) and therefore can be called an 'ontology' , or at least an ontology partition according to me. A nice definition for ontology to which the content of the article conforms is:
"Ontology is an explicit specification of a conceptualization, where the term 'Conceptualization' refers to an abstract model of phenomena in the world by having identified the relevant concepts of those phenomena." Trying to classify cars is a nice attempt to build a small partition of an ontology. [ontology in computer science: LinkBase as a reference - framework = BFO]. Daviddecraene 12:16, 15 June 2006 (UTC)

In computer science the term ontology is indeed used very broadly and sometimes used improperly from a philosophical perspective well that's the nub of the issue: the term means something different in computer science than it does in philosophy. So from a philosophy perspective it is indeed an incorrect definition, but from a computer science perspective its right. Its rather like saying that from a stone-cutting perspective the use of the term "mason" to mean a member of a secret society is improper. This is correct, but the stone-cutting context is not the only valid one. Words can have more than one meaning because they can have more than one context.

I agree that a computer science ontology can also be an ontology in the philosophy sense, I didn't mean to imply otherwise. However it doesn't have to be, which is why we need a separate definition. The term "ontology" in computer science has two important distinctions from its philosophical use:

An ontology is typically what you describe as an ontology partition - it is a model ("specification of a conceptualization") of a specific domain or sub-domain oftenfor some specific purpose. The purpose is often something like "allow users to find answers to queries about Lymphatic Cancer".
An ontology includes the specific instantiation of the model, not just the abstract model itself. In other words I can point to the physical database that contains the ontology.

Hopefully the article captures that. A computer science ontology is obviously related to but not the same as the philosophical definition of "ontology". Gwernol 15:37, 15 June 2006 (UTC)

Perhaps the term "computer science"should be generalized to "computer science" and "information science". Without getting involved in technicalities, one can design computers without any knowledge of ontology, but one can not design information systems without it. DGG 05:07, 9 October 2006 (UTC)

We can argue up to the end of our lives about the distinction, but I would like to deal with facts instead of blah-blah-blah. Each term has an author. As far as I'm concerned the author of term 'ontology' in the field of AI is Tom Gruber with his "specification of a conceptualization" which means "formal description of some conceptual model". We can formalize our representations and understanding of the world in different ways with different instruments like DB schemas, UML, is-a/part-of hierarchy, etc. The fact is, that there is no ideal instrument (language or visualization) to use. This choice is restricted by some context (including the pursued goal). From that point of view both the article and its author's comments are very arguable. Kazmikh 15:54, 15 January 2007 (UTC)

Construing the so-called philosophical sense as "the study of the category of existence" or "of being" looks to me like somebody just looked it up in Webster's; it is an accurate construal of the pre-1950s use of the term in philosophy -- but since the "linguistic turn" in philosophy starting with Wittgenstein and Quine and others, the term has been used in other ways, including a taxonomical delineation of entity types in formal logic. This is why Gruber's definition is *not* different: "a formal specification of a conceptualization" is exactly what you find in a lot of Tarski, Quine, Frege, Kripke -- admittedly not as broad as Cyc or Sumo, but nonetheless very formal. "Description logics" in CompSci are based consciously, admittedly, unashamedly and directly (with footnotes) on the first order logics worked out by philosophers decades earlier for the purpose of expressing their ontologies. So it is easy to argue for the articles being collapsed. That being said, they don't have to be collapsed, it's simply an editorial choice. What would be misleading would be to give the reader the impression that these are mostly different, when in fact they are *completely* overlapping and the only diference is emphasis: computer scientists mainly build actual ontologies and only seldom theorize about them; philosophers often theorize about the right way to build ontologies but rarely build other than very small ones for illustration. Doug Lenat told me to my face he hires as many philosophy PhD's as he can because they are the best, overall, at building Cyc -- so the biggest ontology project in Computer Science is getting built by philosophers. User:tmusgrove

[edit] And Where is the Schema in the Scheme of Things?

Definition of Ontology in this context is much like a Schema (entities and relationships). Upper Ontology is much like a Universal Schema. Comments please. Is there a difference?--Connection 18:33, 5 June 2007 (UTC)

[edit] Example relationship doesn't compute

For example in the domain of automobiles, we might define a made-in relationship which tells us where each car is built. So the Ford Explorer is made-in Louisville. The ontology may also know that Louisville is-in Kentucky and Kentucky is-a state of the USA. Software using this ontology could now answer a question like "which cars are made in America?"

This example doesn't seem to compute. First, how would a computer equate "USA" with "America"? Second, since the query is based on is-in relationships, and the Kentucky/USA relationship is defined "is-a", how would software using this ontology successfully answer this question? At the very least, wouldn't there be some ontological class of 'state' that mandates it has an is-in relationship? I may be picky here, but my understanding is that ontologies need to be painfully complete to support this sort of software application, and examples that don't reflect this are misleading.

Agreed. This seems problematic in a bunch of ways, partly because it treats formal relations as though they were English sentences with commonsense interpretations, partly because it doesn't talk about the many assumptions it is making. For example, what exactly does "A made-in B" mean? Does it mean that *all* A's are made in B, that *some* A's are made in B? Does it mean that they are *currently* made in B? (and what does "currently" mean?) That they have at some point been made in B? That anything that is A is necessarily made in B, the way Roquefort made-in France? (If it's not made in France, it cannot be Roquefort.) What exactly is the predicate in "Kentucky is-a state of the USA"? Is it "is-a"? If so, what is a "state of the USA", and what is its relation to the concept "state" and the concept "USA"? Is it "is-a-state-of"? Then I guess we need (A is-a-state-of B => A is-in B) and (A is-in B & B is-in C => A is-in C). And so on. --Macrakis 21:00, 7 November 2007 (UTC)

[edit] Containing vs. subsuming request for clarification

As somewhat of a novice to the subject... I find these two phrases under the "Classes" section to be confusing and contradictory:

"Ontologies vary on whether classes can contain other classes"
"Importantly, a class can subsume or be subsumed by other classes."

Is there a clear difference between containing and subsuming? If so, I think the article should explain.

Serrin.ling 09:19, 22 October 2007 (UTC)

[edit] Weaknesses of Ontology

Anyone get the feeling here that ontologies are mostly nonsense? An ontology is just an "index" that organizes whatever you are classifying. Without an index, the data are mostly useless, but the index itself is not the data.

And each person would create a different indexing scheme, whatever they might want to call it and however obtuse they make it.

My experience as a webmaster and searching the internet is that you just have to do broad searches and sift through the raw data yourself. There is no magic ontology, index, or omnipotent short-cut. Just grind it out.

Any web page that tries to insert a "catagorization" scheme in front of their data, has just eliminated their data from searches. It might as well not exist.

Before you completely dismiss this comment, there is an analogy to World War II. The allies spent an enormous effort on "pin-point" bombing of the german war factories. The strategy that ultimately prevailed was "carpet bombing", just bomb it all so you don't miss anything.

All this "pin-point" indexing is not nearly as effective as just putting all the data out there and let the search machines have at it.

Thank you! —Preceding unsigned comment added by 134.67.6.15 (talk) 14:54, 30 October 2007 (UTC)

WP Talk pages are not Usenet; they're specifically intended for discussing the page, not the topic. Please move this discussion to your blog or whatever. --Macrakis 15:46, 12 November 2007 (UTC)