Course talk:CPSC522/Ontology

I just skimmed through, (not an accurate feedback). Lack of formalism was the main problem I noticed, that is most of the article is written in words though that might be because of the nature of the topic. Lack of examples and figures is also noticeable but probably not the main problem. The key concepts explained can be considered as the formal contribution.

Since I didn't put enough time to go into the details I can't go too low if I want to score. But the actual quality can be higher or lower.

I think you are right in that I have this "flow" in my mind: concept formation --> formalism --> elaboration or revision. The part of the topic I am interested in is so-called conceptualization and how it leads to formalism. So, I am not starting in the second stage. Also, it might be said that generally we express our thoughts in words; we need natural languages :)

Sorry, I read it more carefully the text was good. Still, it could use more figures and formulas to be more concise but it conveyed the required information.

After taking some time off from the editing and reading the entry again, I realized how unclear my exposition and how bad my writing was in the last draft. I want to express my gratitude (and apologies) to the people who read the earlier drafts and provided feedback to me, especially to Nam and the professor who read the first draft, which was horrible. I hope that the current version is much clear. Shunsuke

Hi, Here are my comments from a read through:

I think the abstract spends too much time talking about the philosophical definition - e.g. the second sentence seems unnecessary.

This sentence is really hard to follow: "In connection to logical reasoning in particular, although this entry is not about ontology as in philosophy, an issue of philosophical nature, the relation of representation of things in the world to the notion of truth, is briefly considered." Is it necessary to draw the connection to philosophy? (ie could you remove "although this entry... ...issue of philosophical nature"? ).

I won't copy and paste them all, but there were other very long sentences throughout the article that I thought would be clearer if they were broken up.

Also in the interests of making things clearer to read and breaking up the text, I thought you could consider using bullet points in some places - like for when you list the essential characteristics of ontologies. Maybe also when listing the three constructs of OWL. For me, at least, this would help to emphasize these points, and make the text easier to follow.

I like how you structure 'Implication 1: Interoperability' and 'Implication 2: Levels of Abstraction' as separate subsections. Maybe more of these mini sections would be better than the bullet points I mentioned above.

Diagrams would be very nice to illustrate the examples you give - for the residential buildings one and/or the Fido one.

In the last paragraph, you use 'Poole says' or 'Poole explains' etc. several times - could you leave these phrases out since you are already citing him? This could help shorten the sentences and make it easier to follow. I think you could maybe do something similar in earlier paragraphs where you reference authors as well.

The topic of ontology somewhat verges on philosophy. The word "ontology" is originally from philosophy; there is a need for disambiguation. It would become apparent if you browse Smith's paper I cited or visit this site mentioned at the CPSC 522 course page: http://ontology.buffalo.edu/smith/

Regarding use of bullet points, personally I prefer to use sentences for precision and clarity. In presentation slides or other kinds of writing, which are essentially notes, I would use bullet points. Generally speaking, pictorial representation does not replace explanation in words. For the Fido example, please see my reply to Nam below. Part of the reason might be that as a former arts minor, I am used to dense prose and tend not find sentences too long. But I will certainly try to mix shorter sentences. Also, a brief summary sentence at the end of a section might help.

As for the last paragraph, generally I prefer to clearly indicate which part is from external sources to avoid plagiarism. But again, I can try to revise wording a little.

The abstract should tell us what an ontology is and why it is important and relevant. The reader should be able to say what an ontology is from reading the abstract. The abstract should be about ontologies, not about the writing of the article. The title should be more like a title; at the moment, it looks more like an abstract (in that it tries to define an ontology) than the abstract itself.

The reader should be able to come away with some idea of what an ontology is. (Or if there are multiple senses of the word used in different communities, what some of these definitions are.) It would be nice to have a clear definition of an ontology. For example, if you were writing about "pens" it would be good to have a definition (e.g., "a pen is a writing instrument that puts ink on a paper or other surface"). There might be multiple definitions, but in this case it might be better to give the multiple definitions rather than not giving any clear definitions. Think about the aim that a reader should be able to explain what an ontology is after reading your article.

It is not clear to me why truth is relevant. (You talk about truth, assuming it is relevant, but not about why it is relevant). When the IAU defined the term "planet", they could have defined it however they liked; there was no notion of a definition being true. We could then argue whether "Pluto is a planet" is true. So if ontologies are about definitions, which you hint at, then why is truth relevant.

You should try to aim this at the intended audience: other CPSC 522 students (present and future). Try to not use terms --- or at least explain them --- that they would not be familiar with. For example, unfortunately (and it is no fault of their's) you shouldn't assume students know the type-token distinction. You should be able to explain it clearly. (I wouldn't use the number 2 as you example, use something simpler like the type dog and a particular dog).

Thank you very much for the feedback. I am saying this not because you are the instructor; your comments actually helped me think about what I wrote. First of all, I certainly know that in metaphysics, ontology is used in a different sense. Disambiguation as well as some form of definition in the computer science community might be useful to give the reader some idea of what this entry is about. The latter is somewhat implicit, however, in the second section; but I can certainly try to state it more explicitly. More importantly, you are right in that truth is not somehow a property associated with the definition or meaning of a term itself. But Smith is concerned about ontology actually being true of the world. I should consider his remark more carefully. A trite example is that a bachelor is an unmarried man. So if Bob fulfills these conditions, the sentence 'Bob is a bachelor' is true. But by the same token, if a definition, be it planet or bachelor, does not capture what we are trying to speak of in the world, it is useless for us. And we would be hard-pressed to say where the notion of truth come from; the talk, done by means of those terms associated with the definitions, would not be about the world any longer. We do need that notion, truth, in formal reasoning. Maybe this is the line I should think about a little more. Needless to say, I will also address the other issues you raised.

Given that I got your feedback only a few hours ago, would it be possible to give me a few more days to incorporate further thoughts in the entry? Writing takes time (because I have to think), and I have impending deadlines for math 419 and marking (I am a TA).

Here is one thing to think about:

We could have the definition "a bachelor is an unmarried man". Call this Bachelor_formal (as it is the formal definition). Assume it is a fine definition (assuming man and unmarried are defined). We could then discuss as to how much natural usage follows this definition (and there are many ways to measure this, usually with two numbers). And when we want to be careful we would specify whether we mean Bachelor_formal or Bachelor_common_usage. In science, it turns out that it is better to be precise, so that we more and more use planet_formal rather than planet_common_usage. Then, when we specify that Pluto is a planet_formal then we can measure the properties of Pluto and come to a conclusion (and perhaps even be unsure because we can't measure all of the properties of Pluto). When we want to determine whether Pluto is a planet_common_usage, then first it changes over time, and secondly it is more a question of sociology than of the properties of Pluto.

(See the Canvas post for new deadlines)

Thank you very much for the additional feedback and letting me know of the post. I will try to think over this question, relating philosophical talks to the scientific context. For now, I have to attend to the math assignment, which is falling apart :)

Hi,

Here's the comments I've had from a single read through your paper. I will add some more as I go on.

Abstract

> In this entry, I first introduce the notion of ontology in such a manner as to not only give a high level understanding of what ontology is and of what practice importance it has but also to appeal to the intuitions of the reader about representation of knowledge.

Wordy sentence.

Introduction

> Let me begin with the following remarks from a philosophical discussion

Not sure first person voice is appropriate for the article. Repeats.

> I duly admit that I am glossing over the subtleties in the original context.

I think disclaimers should be more objective, e.g. "This page may not comprehensively capture the subtleties..."

> science researches

Is "researches" a correct term? I've never seen plral form of research.

Characteristics of Ontology

> which correspond to things construed as types, rather than as tokens.

What's the difference between types and tokens?

> To move on to the second point, namely kinds of and levels of abstraction of terms, the discussion so far may have given the impression that all representation scheme operate at the same level of abstraction with respect to other systems

Wordy sentence.

> Conceptions of Truth: What to Represent

This part does not seem to motivate the formal language expression of ontology. I see the philosophical discussion about truth is very interesting, but it might be too high level for the scope of the article.

> To use an example similar to Russel and Norvig [6], a semantic network might have a node "Fido", which is connected to another node "Dog" by an edge labeled with "MemeberOf", the latter, in turn, to the node "Canidae" by the relation of subset. In particular, the construction is to be expressed in description logic sentences.

An image accompanying this example would be nice.

Connection of Ontology to Probabilistic Models

> My presentation here is only for the purpose of giving a rough idea about how ontology might lend itself to probablistic interpretations and, as such, neither describes his approach in entirety (in particular, I skip the discussion of conditional probability) nor includes all details and subtleties.

Same point about disclaimers and wordiness.

This is not full response to all of your points. Let me reply for now at least to some of the points (sorry, not necessarily in the order of your list).

1) The expression "wordy" is rather subjective; in particular, precision in expressions as well as grammatical considerations, such as parallelism, sometime do require "long" sentences -- for instance, legal documents. To be sure, this is not to say all of my sentences are carefully crafted, and some of them could be reworded, which I will work on. I would add that you should use sentences to explain your views fully; otherwise, it sounds as if you are stating your personal opinions, which I am not obliged to agree.

2) I would say that the discussion about truth is relevant in that ultimately we want to capture truth about the world in our representations of domains. (Does not science have a similar goal?) Note that the relationship between human-devised systems -- be they physics, natural languages, or knowledge bases -- and what things are in the world is not simple; it seems that there is no necessity that such systems somehow fit to things in the world. Concepts and real things are in nature very different. We simply say in logical deduction that a conclusion reached by inference is true or false, but what do you mean by the word "true"? How can you be sure that sentences you use actually reflect what things are in the world and on that ground are true? I want to see the process from conceptualization to knowledge representation and artificial intelligence systems in particular. Realism and correspondence theory of truth, which are expressed in the Smith's remarks, are fundamental for ontology; I would not describe it as "high level" discussion. That being my response, however, I could state more explicitly why I think that it is relevant.

3) I am not sure if the type and token distinction is technical; synonyms for these two words are kinds and instances. When I write number 2 on a blackboard, that is a token of the type number 2. The explanation in terms of type and token is my paraphrase of the original passage. People might say somehting to the effect that "as a token of my gratitude (or appreciation),...".

4) Regarding my use of the first person in the exposition, I know that the convention is to use an objective tone in reports. I would say, however, rhetorical (here in a derogatory sense) words I might use do not make the content any more objective and unbiased. (Just think of glib, pompous talks by politicians or lawyers, with little substance.) One should judge the worth of a report on grounds of the substance, namely the argument and evidence. At any rate, given that I am the sole author, whose opinion could it be if not mine?

5) The word "researches" is actually correct. See for instance https://dictionary.cambridge.org/dictionary/english/research It says that is a British convention. But to my ears, generally words seem to have a different nuance when used in plural.

6) I think that you might be right about graphical representation, generally speaking; there is a figure in a paper I was thinking to add to give an idea about what ontology might look like. However, insofar as that particular example is concerned, I do not think a pictorial representation adds anything. I explicitly say there are nodes and edges in the graphical representation. In fact there are only 3 nodes. The relations of membership and subset can go only in one direction, so that edges are directed. Fido is a dog's name. So Fido is a dog. Canidae is a family, to which dogs belong. So, you know which direction the edges point. I am not sure what is not clear here, to be honest. What is the ambiguity you are concerned about?