W3C home > Mailing lists > Public > semantic-web@w3.org > February 2009

Re: live meaning and dead languages

From: Hugh Glaser <hg@ecs.soton.ac.uk>
Date: Mon, 9 Feb 2009 12:07:50 +0000
To: Charles Dylan Shearer <dshearer@nekonya.info>
CC: Tim Berners-Lee <timbl@w3.org>, Jeremy Carroll <jeremy@topquadrant.com>, "semantic-web@w3.org" <semantic-web@w3.org>
Message-ID: <EMEW,l18C810459f29a51026fb0dc77f9742a388f51,hg%ecs.soton.ac.uk,C5B5CD16.29479%hg@ecs.soton.ac.uk>

One insertion below.
On 09/02/2009 02:38, "Charles Dylan Shearer" <dshearer@nekonya.info> wrote:

> Hugh, you provided an interesting example.  I think we should make
> some observations about the difference between words and the things
> they denote, using your example as illustration.
>
> The question of this discussion is what implications does the fact
> that the meanings of words in natural languages change have for the
> long-term utility of the Semantic Web's method of vocabulary
> creation.  You alluded to the fact that, in the 1960s, the word
> "hacker" meant something like "a person who programs as a hobby", but
> in the 80s its began to mean something like "a person who tries to
> gain unauthorized access to networked computers".
>
>> For example, if there had been a SW in the 1960s, there would have
>> been a URI for the concept addressed by the word "hacker", say,
>> http://dbpedia.org/resource/Hacker
>>  or even http://tmrc.mit.edu/dictionary.rdf#hacker.
>> However, once it had appeared in the metadata for TRON, and War
>> Games, or whatever, in the 1980s, the URI http://dbpedia.org/resource/Hacker
>>  would have changed its meaning, with the most widespread meaning
>> still being that usage today.
>> And no amount of pointing the world at dbpedia and other definitions
>> would ever change that.
>> So "meaning as use" is not the basis of the design, but it is a
>> reality we must face in the long term.
>
>
> You argue that changes in words' meanings will cause a problem for the
> Semantic Web, by arguing that the change in the meaning of "hacker"
> would cause a problem for its related URI (let's pick
> "http://dbpedia.org/resource/Hacker
> ").  But I think that it is not clear what the URI
> "http://dbpedia.org/resource/Hacker
> " refers to.  You said that "http://dbpedia.org/resource/Hacker" is a
> URI for the concept addressed (in the 60s) by the word "hacker".  If
> that were so, then the change in the meaning of "hacker" in the 80s
> should have no impact on that URI: the change involved the *word*
> "hacker", not the concept "a person who programs as a hobby".  The use
> of the word "definitions" in your comment "no amount of pointing the
> world at dbpedia and other definitions would ever change that" betrays
> your switch, at the end of the paragraph, to assuming that the URI
> refers to the word, which is contrary to your earlier statement.
> After the 1980s, the world had two concepts: "a person who programs as
> a hobby" and "a person who tries to gain unauthorized access to
> networked computers".  The word "hacker" had been changed from
> pointing to the first concept to pointing to the second, but I think
> that both concepts were well-used in computerized societies (the first
> one was more often referred to with the words "nerd" and "geek"
> instead of "hacker").  So, I think that we can conclude that, if
> "http://dbpedia.org/resource/Hacker
> " is the URI of the concept, then the change in the word "hacker" has
> no impact on it.
>
> As I pointed out, you seemed to draw your conclusion from the
> assumption that "http://dbpedia.org/resource/Hacker" refers to the
> word "hacker" (rather than a concept), but I believe that, even in
> this case, the change in the meaning of "hacker" does not cause a
> problem for that URI.  Because the word "hacker" changed, it would be
> proper to modify the published assertions involving the URI
> "http://dbpedia.org/resource/Hacker
> " to reflect these changes, just as an English dictionary would modify
> its entry for "hacker".  If the word "hacker" had become ambiguous,
> then presumably the document retrieved by dereferencing
> "http://dbpedia.org/resource/Hacker
> " would mention that ambiguity.  The important point is that
> "http://dbpedia.org/resource/Hacker
> " itself did not become ambiguous -- it still referred to the word
> spelled "h"+"a"+"c"+"k"+"e"+"r".
I was not intending to deal with words at all - sorry if I misled.
I was always talking about the URI http://dbpedia.org/resource/Hacker.
To start with this may be agreed to be the concept "a person who programs as
a hobby" (we won't go into how that agreement was reached).
If this URI then gets used in a context where it is no longer the concept "a
person who programs as a hobby", as it would have in the RDF metadata for
the films (perhaps), where does that leave the URI?
Does the dbpedia RDF provider decide to conform to the changed meaning (as a
dictionary might), or not?
Clearly they have a choice.
If they have a choice, then it is an interesting problem for us to study.
A) Change - dbpedia probably would, because it aims to be like a dictionary.
B) Not change - I am annoyed I have lost the original meaning of the
concept, and continue to use it in the RDF versions of my lecture notes,
without acknowledging (ie referring to) the new meaning (not really, but
could be, and is true of the use of the word "engineer" by the professional
bodies in the UK).

In the end, I don't even "own" the URIs I mint, even if I own the domain.
These are interesting socio-technical issues, I think.
>
> There is one alternative meaning of the URI
> "http://dbpedia.org/resource/Hacker
> " that would cause the change in the meaning of "hacker" to change
> what "http://dbpedia.org/resource/Hacker" *ultimately refers to*.  In
> this case, "http://dbpedia.org/resource/Hacker" means "the meaning of
> the word 'hacker' " -- that is, it refers to a specific concept only
> when that concept happens to be the meaning of the word "hacker".  So,
> when the meaning of "hacker" changed, the referent of
> "http://dbpedia.org/resource/Hacker
> " changed accordingly, but not the *meaning* of
> "http://dbpedia.org/resource/Hacker
> ".  Again, I think that the change in "hacker" does not create a
> problem.
>
> In general, I think that ontologies on the Semantic Web will be most
> affected by changes in the concepts people use rather than the changes
> in word-meanings that often accompany them.
>
> Dylan
>
>
> On Feb 8, 2009, at 9:19 AM, Hugh Glaser wrote:
>
>>
>> Hi.
>> I expect it to be the case that meaning can seriously change, and
>> there is an analogy with natural language.
>> (URIs based on formal groups such as WGS84 will always tend to be
>> more stable, of course, just as words are.)
>> For example, if there had been a SW in the 1960s, there would have
>> been a URI for the concept addressed by the word "hacker", say,
>> http://dbpedia.org/resource/Hacker
>>  or even http://tmrc.mit.edu/dictionary.rdf#hacker.
>> However, once it had appeared in the metadata for TRON, and War
>> Games, or whatever, in the 1980s, the URI http://dbpedia.org/resource/Hacker
>>  would have changed its meaning, with the most widespread meaning
>> still being that usage today.
>> And no amount of pointing the world at dbpedia and other definitions
>> would ever change that.
>> So "meaning as use" is not the basis of the design, but it is a
>> reality we must face in the long term.
>> Of course, this is not just a problem - it is an exciting challenge!
>> And the fun is to work out how the SW can give us a better handle on
>> all this.
>>
>> <cod philosophy warning>
>> I consider that there is a great danger when designing systems that
>> they get built for the way we would like the world to be, not the
>> way it is. That way lies failure (cf the attitude to broken links in
>> the old hypertext world already mentioned).
>> </cod philosophy warning>
>>
>> Best
>> Hugh
>>
>> On 08/02/2009 00:12, "Tim Berners-Lee" <timbl@w3.org> wrote:
>>
>>
>>
>> Jeremy,
>>
>> The Semantic Web is a system we design, not one we observe from the
>> sidelines as we do with Latin, say, or English.  It is simply a
>> different system. While analogies with natural languages are sometimes
>> useful they are often quit misleading. In particular, meaning as use
>> is not the basis of the design.
>>
>> The design of HTTP space is that URIs -- terms in the language --  are
>> owned by particular agency -- people or communities of them.
>> The terms are defined well enough for a particular use.
>> The defining agency is often available to resolve the meaning n the
>> case of dispute.
>> The lifetime over which the terms are useful are the lifetime of the
>> application.
>>
>> For example:
>>
>> wgs84:lat is short for a URI defined by the international group WGS84.
>> The rdf property makes reference to a spec which has been implemented
>> thousands of times.
>> It is is well known as the numbers you get out of a GPS unit.
>> The term can only be used for positions on earth in a particular
>> projection.
>> If a new GPS system comes into play in the future which is much better
>> than the current, and uses a different projection, then new GPS
>> systems may be built to use new different terms by default.
>> The new terms will of course use a different URI, so there will be no
>> confusion.
>> But the relationships  between the projections will be well
>> documented.
>> And the WGS84 system I would expect to have a very valuable life, huge
>> amounts of data being reliably exchanged using it, fro maybe
>> centuries, and I hope even in future millennia being a dead language
>> which is extremely well known by future librarians (human or not).
>>
>> This stability of terms like this if very important for the systems we
>> are building.
>> You say you are an engineer not a philosopher.
>> I would, as you know, call you a philosophical engineer.  We are, in
>> the semantic web, building systems in which we engineer the philosophy
>> so that the system will have properties we desire.
>>
>> Tim
>>
>> On 2009-02 -05, at 22:34, Jeremy Carroll wrote:
>>
>>>
>>>
>>> One of the occasional defects of people in SW is a tendency to arm
>>> chair philosophizing.
>>> I will indulge.[...]
>>
>>
>>
>
>
Received on Monday, 9 February 2009 12:08:45 UTC

This archive was generated by hypermail 2.4.0 : Tuesday, 5 July 2022 08:45:10 UTC