W3C home > Mailing lists > Public > www-rdf-comments@w3.org > April to June 2002

Re: I'm Confused About RDF: A Subject is Not A Predicate, etc.

From: patrick hayes <phayes@ai.uwf.edu>
Date: Mon, 13 May 2002 01:02:55 -0500
Message-Id: <a05111703b904d14d6c94@[]>
To: "Ivon Fergus" <ivonfergus@hotmail.com>
Cc: www-rdf-comments@w3.org
>First, I need to do a lot more research before I get into the 
>subject of RDF at any length. However, because the magazine New 
>Architect (formerly Web Techniques) has published an article in its 
>June 2002 edition by Dr. Uche Ogbuji entitled "The Languages of The 
>Semantic Web", I feel that it is vital to get the ball rolling now, 
>even though I'm not at all sure in which direction. Specifically, on 
>p. 31, Dr. Ogbuchi makes the following amazing statement:
>"RDF statements are hardly as complex as those we use in natural 
>language. They have a uniform structure of three parts: predicate, 
>subject, and object. For example,: the author [predicate] of The 
>Lord of the Rings [subject] is J.R.R. Tolkien [object]."
>I will make some assumptions; Please correct me if I am wrong.  My 
>first assumption is that there must be a valid reason to totally 
>diverge from the common meanings of these terms.

Unfortunately there is no 'common meaning' for these terms. What 
meaning they have depend on ones technical background and the 
assumptions made by the particular field one works in. Certainly, 
however, one should not think of RDF in terms developed for 
linguistic analysis of natural language.

>  For example, the word "author" is the subject of the sentence, not 
>the predicate.

You may have found Ogbuji's example less confusing if it had been 
phrased thus: "The Lord of the Rings [subject]  has as 
author[predicate] J.R.R. Tolkien [object]" , retaining the triple 
ordering in the English syntax. However, I presume his rendering was 
intended to convey the intended meaning rather than the grammatical 
structure. The RDF terms refer to the 'subject', 'predicate' and 
'object' of the RDF triple, not of the explanatory sentence. (The 
exact terminology is not in fact terribly important: they could be 
called 'middle/first/last' and the content of the RDF would not be 
altered in the slightest degree. The terms chosen were intended to be 
intuitively helpful, in that an RDF triple is often thought of 
intuitively as being 'about' its subject, but if you find the 
intended (weak) analogy unhelpful, then it would be best to ignore it 

>The booktitle is the object of the preposition "of", not the 
>subject, and only in some respects can "J.R.R. Tolkien" be 
>considered an object.When I went to school, the noun or pronoun 
>following an intransitive verb, such as "is" was called a predicate 
>nominative, not an object. For example on web page: 
>is the following statement:
>"Predicate Nominative
>A predicate nominative is a noun or pronoun which follows the verb 
>and describes or renames the subject. It is another way of naming 
>the subject. It follows a linking verb. The predicate nominatives on 
>this page have been italicized."
>I assume that "linking verb" is a synonym for intransitive 
>verb.Perhaps a few things have changed a bit since I was in school, 
>but the only time we referred to a noun or pronoun as an object was 
>when it followed a transitive verb and was its object.

The topic of your schooldays was apparently English grammatical 
analysis of a certain traditional kind (used in 'prescriptive' rather 
than 'scientific' linguistics.)  RDF is not English, however. It is 
closer to the kinds of information found in relational databases than 
it is to natural language.

It might be better to think of an RDF triplet as asserting that a 
relationship holds between two entities. The above example then 
asserts that a relationship called "author"  (or possibly "author_of" 
)  holds between the entity called "The Lord of the Rings" and the 
entity called "J.R.R.Tolkein", in that order. This could be expressed 
in English in a number of grammatically distinct ways; the 
'predicate/subject/object' terminology is not meant to refer to the 
grammatical structure of those English sentences, however.

>For example in the sentence, "He bought a book." the noun "book" is 
>the object because it has transitioned in some way.  In this case, 
>certainly not physically, but in terms of ownership.

Speaking now as a logician, I would argue with the coherence of this 
characterization. This way of talking seems to me to be rooted in a 
rather primitive (Aristotelian) view of logical structure, which was 
superceded by mathematical logic in the late 19th/early 20th century. 
Why is it correct to say that when I buy a book , that the book has 
'transitioned' but I have not? An ownership property of the book has 
indeed changed, but it is also correct to say that an ownership 
property of me has changed: I now own a book that I did not own 
before. It is more accurate to say that an event of buying is a 
change in a *relationship* between two things (me and the book). The 
subject/object distinction is in fact a linguistically arbitrary 
aspect of English grammar, and has no logically coherent 
justification. The same event can be described in the English passive 
with the book as the grammatical subject: "The book was bought by me."

>In the sentence, "He burned a book." it is more obvious that the 
>book is the object, and English grammar also defines indirect 
>objects as well as direct objects, so I can understand that 
>reasonable people can disagree about exactly what is a certain part 
>of speech and quibble in the finer parts of English grammar.

There are (at least) as many English grammars are there are theories 
of linguistic structure. None of them are universally accepted or 
definitive, however, so it seems to me to be meaningless to refer to 
"English grammar" in the singular.

>Nevertheless, Dr. Ogbuchi's rendition of RDF grammar in this passage 
>is so disjoint from "natural language" as to be incomprehensible by 
>an educated person.

The basic mistake is to try to read the RDF as though it were natural 
language. Ogbuji was trying to be helpful, but I think you read far 
too much into his example.

>I am at a complete loss as to why these definitions depart so 
>radically from what I learned in school. I decided that there must 
>be very important consideraitons that I am oblivious to that 
>necessitated this massive shift, or perhaps Dr. Ogbuci or New 
>Architect needs to issue a clarification to this article.

I am not sure where on when you went to school, but these ideas 
certainly pre-date my own school years. The  basic 'massive shift' 
took place in the work on relations by Peano,  Frege and others 
towards the end of the 19th century, and was made permanent in the 
English-speaking world by "Principia Mathematica" by Russell and 
Whitehead , published in 1910.  Almost all of logical semantics has 
been based on these ideas since the work of Tarski in the 1940s, and 
modern linguistic theories have been predicated on similar underlying 
semantic and grammatical structures since the work of Chomsky in the 
1960s, and possibly earlier.

>Having deduced that much, I went to some of the relevant draft 
>documents on the W3 website, seeking enlightenment. I briefly 
>reviewed <http://www.w3.org/RDF/>http://www.w3.org/RDF/ , 
><>http://www.w3.org/TR/rdf-mt/ , and <>http://www.w3.org/TR/rdf-mt/, 
>and now must confess that I am even more profoundly confused than 
>before. Now, I am not an ignoramous, so I beg for the RDF primer 
>that one of these documents alluded to, so that I can begin to 
>understand these documents and thus RDF grammar, syntax, and 
>effective usage.

In a nutshell: think of the RDF as a graph. The nodes of the graph 
name entities, and the arcs of the graph refer to relationships. The 
RDF graph asserts that relationships hold among the entities in the 
pattern defined by the graph. (Other relationships may also hold 
among them; the RDF graph does not deny that possibility.)   A 
'blank' node in a graph asserts that an entity exists without 
actually naming it, as for example when one might wish to say that 
someone has a son, without knowing the son's name.

Exactly what counts as an entity is left open, so that RDF can 
describe relationships among arbitrary entities; similarly, exactly 
what counts as a relationship is left open; all that RDF requires of 
a relationship is that it hold or not hold between entities; in other 
words, that RDF triples are either true or false.

I hope this has been of some help.

Pat Hayes
Received on Monday, 13 May 2002 11:32:41 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:15:18 UTC