concern about label 'is_about' for 0000115
Jonathan Rees
<obo:IAO_0000115 rdf:datatype="http://www.w3.org/2001/
XMLSchema#string"
>a representation is_about anything exactly when some part
of the information_artifact denotes the entity</obo:IAO_0000115>
I am still uncomfortable with this. The definition is perfectly clear,
but it does not match the way 'about' is used in ordinary language. I
think the definition matches the word 'mentions' better than it
matches 'is_about'. Consider:
- an information artifact can be about something, without mentioning
it (as Chronicles of Narnia are about a particular deity without
mentioning it, or the string 'Weight: 52 kg' can be about a particular
person, without mentioning the person)
- an information artifact can have a part that denotes something,
without being about it (e.g. the statement 'Samuel drove through
Cleveland to get to Detroit' mentions Cleveland without being about
it. It might be about Samuel, but it says nothing whatsoever about
Cleveland)
Can we consider relabeling this definition 'mentions' instead of
'is_about', so that some future theory can tell us what 'is_about'
might really be about? Otherwise I fear people may be misled by the
label into reading too much into assertions that use this relation.
Alan Ruttenberg
condition for aboutness. I appreciate the contrary examples of
aboutness. My preference would be to tighten up the definition rather
than change the label. In particular we have subclasses of is_about
that are more specific. The one 'is quality measurement of' is exactly
of the sort that you imply with your '52 kg' example.
Do you have more examples of non 'mentioning' aboutness that we can
record? Any clue as to necessary parts of the definition?
-Alan
Jonathan Rees
<alanrut...@gmail.com> wrote:
> Do you have more examples of non 'mentioning' aboutness that we can
> record? Any clue as to necessary parts of the definition?
I'll think about this but that second question is extremely difficult
- if I could answer
it I'd be a better philosopher than Brian Cantwell Smith, who has been
struggling with it for decades, and I'm not.
For the purposes of IAO, if you want to go this route, there might be
alternatives
to a principled answer. For example, one could give lots of examples of the sort
I gave (and that I will try to come up with). Or one could try to get
users of the
ontology to induce intended correct practice from given use cases
or curation needs; that would be somewhat better.
Perhaps the library science profession has something to say on the
question, as 'is about' may
be very similar to the notion of subject heading. A book that mentions Cleveland
only in passing will not acquire 'Cleveland' as a subject heading, while a book
that has a chapter about Cleveland will. Perhaps one learns in library science
school when it's good practice to apply a subject heading and when
it's not. If that
principle is articulated maybe we could adapt it.
Alan Ruttenberg
>
> On Tue, Mar 24, 2009 at 9:02 AM, Alan Ruttenberg
> <alanrut...@gmail.com> wrote:
>> Do you have more examples of non 'mentioning' aboutness that we can
>> record? Any clue as to necessary parts of the definition?
>
> I'll think about this but that second question is extremely difficult
> - if I could answer
> it I'd be a better philosopher than Brian Cantwell Smith, who has been
> struggling with it for decades, and I'm not.
> For the purposes of IAO, if you want to go this route, there might be
> alternatives
> to a principled answer. For example, one could give lots of examples of the sort
> I gave (and that I will try to come up with). Or one could try to get
> users of the
> ontology to induce intended correct practice from given use cases
> or curation needs; that would be somewhat better.
>
> Perhaps the library science profession has something to say on the
> question, as 'is about' may
> be very similar to the notion of subject heading. A book that mentions Cleveland
> only in passing will not acquire 'Cleveland' as a subject heading, while a book
> that has a chapter about Cleveland will. Perhaps one learns in library science
> school when it's good practice to apply a subject heading and when
> it's not. If that principle is articulated maybe we could adapt it.
suggest it is about?
I think being the subject of something is narrower than being about.
It is more like foaf:primaryTopic.
Another sense of aboutness that the current definition supports is
more along the lines of what you had discussed at some point as a
definition of information entity: That which informs.
-Alan
>
> >
>
Barry Smith
>- an information artifact can have a part that denotes something,
>without being about it (e.g. the statement 'Samuel drove through
>Cleveland to get to Detroit' mentions Cleveland without being about
>it. It might be about Samuel, but it says nothing whatsoever about
>Cleveland)
I think this statement is about Cleveland. That this relation obtains
may not be so obvious to the average speaker of English, but it is, I
believe, a coherent relation, and of the right generality for IAO.
BS
Alan Ruttenberg
On Tue, Mar 24, 2009 at 8:52 AM, Jonathan Rees <jonath...@gmail.com> wrote:
>
> Just a detail...
>
> <obo:IAO_0000115 rdf:datatype="http://www.w3.org/2001/
> XMLSchema#string"
> >a representation is_about anything exactly when some part
> of the information_artifact denotes the entity</obo:IAO_0000115>
>
> I am still uncomfortable with this. The definition is perfectly clear,
> but it does not match the way 'about' is used in ordinary language. I
> think the definition matches the word 'mentions' better than it
> matches 'is_about'. Consider:
>
> - an information artifact can be about something, without mentioning
> it (as Chronicles of Narnia are about a particular deity without
> mentioning it, or the string 'Weight: 52 kg' can be about a particular
> person, without mentioning the person)
Here a question to ask would be whether it is the intention of the
author that when they used the term "Aslan" they intended to denote
that other deity. IAO is based on the principle that it is the
originator of the information that is responsible for its meaning,
part of which includes denotation.
I do think, however, that something needs to be done so that
expression of the definition is such that it is considered sufficient,
but not necessary.
-Alan
Jonathan Rees
between "mentions" and "is about" is not wasted. If you mean
"mentions", call the relation "mentions," per the current definition.
If you mean something different, provide a definition that reflects
this difference.
So far neither you nor Barry has provided any example where you mean
anything more than "mentions". If there is no such example then forget
about "is about" as it misleadingly sounds stronger and deeper than
"mentions".
I don't have a better definition of "is about" to offer, just the
intuitions I've previously conveyed.
Jonathan
Dan Brickley
Just for compare/contrast...
In FOAF, we define topic (and an inverse, page), also primaryTopic.
http://xmlns.com/foaf/0.1/#term_primaryTopic
"The foaf:primaryTopic property relates a document to the main thing
that the document is about. "
http://xmlns.com/foaf/spec/#term_topic
"The foaf:topic property relates a document to a thing that the document
is about."
We don't in FOAF give a name for "mentions", but the existence of
"primaryTopic" frees up foaf:topic to be used more generously / loosly,
eg. for relatively passing mentions.
There is real value in primaryTopic, since it allows document
identifiers to be used as indirect identifiers for the things they're
primarily about. Of course, a primary topic isn't clear for each and
every document, but that's ok.
From the FOAF spec, "The foaf:primaryTopic property is functional: for
any document it applies to, it can have at most one value. This is
useful, as it allows for data merging. In many cases it may be difficult
for third parties to determine the primary topic of a document, but in a
useful number of cases (eg. descriptions of movies, restaurants,
politicians, ...) it should be reasonably obvious. Documents are very
often the most authoritative source of information about their own
primary topics, although this cannot be guaranteed since documents
cannot be assumed to be accurate, honest etc. "
cheers,
Dan
Jonathan Rees
not sure the FOAF definition helps IAO, since it just says that "x is
a topic of y" is defined to mean "y is about x" which is not really
very helpful guidance to a curator. I suppose we could simply define
"y is_about x when y is about x (and/or x is a topic of y)" and hope
for the best, but I understand that the OBO Foundry approach strives
for tighter definitions than this. Perhaps this is a case of
diminishing returns... [I say this in order to provoke Alan]
Assuming we have agreement (or agreement not to pursue further) on "is
a topic of" or more likely "is_about", it may be useful for IAO, like
FOAF, to have an additional relation "is the primary topic of" or "is
primarily about".
By the way the word "topic" will be about as welcome as "concept" in
OBO-Foundry-land - am I a topic? what is not a topic? - but "is about"
clearly has very similar intent to faof:topic.
Jonathan
Paolo Ciccarese
which is something interesting for the work we are doing with the SWAN
ontology. What follows is more brainstorming material than the
expression of a strong opinion and might be a bit confused, but I am
trying anyway.
SWAN intro
----------------
As you might not be familiar with the SWAN project and ontology
( http://swan.mindinformatics.org/ontology.html ) I'll try to briefly
summarize our typical use case. In our SWAN beta application we have
currently some thousands of scientific statements. These can be
research statement (claims or hypothesis), comments or research
questions. The scientific statements are curated through different
processes but the main sources are actual journal articles. Our
curators read the articles and break them down in a list of scientific
statement representing the scientific discourse expressed by that
published contribution.
Use case
-------------
To give a better idea, a SWAN scientific statement, in this case a
claim, can be something like: "The primary function of the MAP tau,
which is particularly abundant in the axons of neurons, is to
stabilize microtubules (MTs)".
http://hypothesis.alzforum.org/discourseelement/134
title: The primary function of the MAP tau, which is particularly
abundant in the axons of neurons, is to stabilize microtubules (MTs)
author http://hypothesis.alzforum.org/person/324
author http://hypothesis.alzforum.org/person/325
author http://hypothesis.alzforum.org/person/326
curator http://hypothesis.alzforum.org/person/1
Existing Annotation
--------------------------
Our curators attached annotation to this particular statement in
three different ways:
1) the statement "refers to" the protein "Microtubule-associated
protein tau [Homo sapiens]". In the SWAN ecosystem the protein has a
unique identifier (URI). The OWL type of that URI is "Protein" and we
might have documents about that protein and they would have different
URIs. We used the relationship "refers to" to express the fact that
the original statement is actually containing a chunk of text that
refers to that protein. There might be the need of some communication
with the original author to figure out the organism the protein is
related to, but still the protein is mentioned in the text.
http://hypothesis.alzforum.org/discourseelement/134
refers_to http://hypothesis.alzforum.org/protein/326
curator http://hypothesis.alzforum.org/person/1
Moreover, we can safely assert that, if the statement is referring to
a protein, the protein is somehow related to that narrative object.
Trying to use "has_topic" and "is_about" we will obtain:
http://hypothesis.alzforum.org/discourseelement/134 has_topic
http://hypothesis.alzforum.org/protein/326
http://hypothesis.alzforum.org/discourseelement/134 is_about
http://hypothesis.alzforum.org/protein/326
I honestly don't see big difference in saying "has_topic", "is_about"
or (the dreaded) "related_to". We know already much more, in fact the
protein was explicitly cited in the text. And even if i project these
relationships to the higher level of the original document. I still
prefer to say that the document "cites" the protein instead of saying
"is_about" the protein. (Unless, that protein is the central topic
according to a curator?)
And I am wondering, given a scientific statement like: Protein A,
unlike protein B, is influencing pathway C. Probably the statement is
about both proteins but the topic is protein A? Both? It is indeed
really subjective.
2) the statement "has_pathogenic_narrative" "initial condition".
http://hypothesis.alzforum.org/discourseelement/134
has_pathogenic_narrative http://hypothesis.alzforum.org/term/initial_condition
curator http://hypothesis.alzforum.org/person/1
This is a classification of the statement through terms coming from a
controlled vocabulary. It is another way of saying the statement
"is_about" but it is usually on a higher level than referring to a
life science entity and usually is not explicitly asserted in the
original statement. Again, we could assert that the "pathogenic
narrative -> initial condition" is a "topic" of the narrative object.
Or the narrative object "is_about" "initial condition". In the SWAN
context we already have a more specialized relationship for saying it.
But this time if I go up at the higher level of the original document
it could be different. I could probably say: the document "is_about"
or "has topic" -> "initial condition". This is because in my mind the
"initial condition" connection wasn't expressed explicitly by the text
but was a form of judgment. Thus, the document doesn't "cites"
explicitly "initial condition".
3) the statement "cites" some evidence (journal articles, comments...)
http://hypothesis.alzforum.org/discourseelement/134
cites_supporting_evidence http://hypothesis.alzforum.org/article/3342
cites_supporting_evidence http://hypothesis.alzforum.org/article/3674
I guess these are not related to the use of "is_about" or "has_topic"
Future Annotation?
--------------------------
We are exploring such as "is_about" more for documents than scientific
statements. But referring to currently, in place, use cases I would
also like to say (referring to the use case statement):
- main topics: "the primary function of the tau protein" or "the
relationship between that protein and the microtubules". I believe it
is a matter of perspectives, SWAN is not like wikipedia where every
document is supposed to be about one topic. Statements can have
multiple main topics. The main topics can be provided by curators or
by the original authors (the distinctive criteria is not who did it
but the nature of it, explicitly expressed or judgmental).
- topics (or about): I would like to be able to say that a scientific
statement is about something resulting from a judgment such as an
editorial decision on what the topics are. When the topics are
explicitly expressed we can still use the relationship "cites" or
something like "refers to". For instance the following are explicitly
mentioned : tau protein function, microtubules, the relationship
between the protein and the microtubules. For these we would use the
two relationships to cover those areas we are not covering already
with more specific relationships. Again, topic (or about) can be more
useful for things that are the result of a judgment.
Basically I would like to be able to:
- have multiple main topics (in SWAN we currently use FOAF, we
appreciated the change in the range/domain of the 'topic' related
properties but the limitation of having a single main topic is a big
one).
- distinguish between what the original author stated and what has
been added as result of judgment.
Given the issues I have described, I am wondering, how would be a good
use of "is_about" in my model?
Paolo Ciccarese