Re: on Use cases/ Web portals

Jim Farrugia wrote:
> Jeff,
> 
> Thanks for your reply.
> 
> Your reworked version clears up the issues in my first three comments.
> It also resolves some of the issues of my fourth and fifth comments.
> 
> The remaining concerns I have are about: (1) the idea of improving
> the quality of search results, and (2) the example inferences you provide.
> 
> The following comments address these issues.
> 
> 1. I would suggest something along the following lines as an
>    alternative to your first sentence. (The idea for me would be, say,
>    to mention both terminology and rules first, then say that once
>    terminology and rules are in place, inferences can be made;
>    and these inferences can somehow "enhance" search results.
>    I think once you start saying "improve the quality of search"
>    you're almost obligated to talk about information-retrieval
>    measures like recall and precision, even if only to contrast your
>    notion of improved search quality with conventional benchmarks of
>    retrieval effectiveness. I'm not sure whether you would want
>    to go down that path with this example, though I think it could
>    be very interesting.  The question marks below indicate my own
>    uncertainty about the wording that I'm suggesting.)
> 
>   This ontology can provide a terminology for describing content,
>   as well as rules that ??describe relationships among pieces of this
>   content??. Given suitable terminology and rules, an ontology allows
>   certain inferences to be made. These inferences can, in turn, allow
>   users to obtain search results from the portal that are difficult
>   to obtain from conventional retrieval systems.

One of the problems here is that we are trying to describe something
without getting overly technical. To make this truly precise, we would
have to provide an introduction to logic and semantics, which we do not
want to do in this document. How about something along the lines of:

This ontology can provide a termionology for describing content and
axioms that define terms using other terms from the ontology. For
example, an ontology might include terminology such as "journal paper,"
"publication," "person," and "organization." This ontology could include
definitions that state things such as "all journal papers are
publications" or "the authors of all publications are people." When
combined with facts, these definitions allow other facts that are
necessarily true to be inferred. These inferences can, in turn, allow
users to obtain search results from the portal that are impossible to
obtain from conventional retrieval systems.

> 2. Your next-to-last sentence says that the rules "allow query systems
>    to infer additional information about journal papers and authors," and
>    your last sentence suggests (if I understand it right) that a search
>    engine could return [instances of] journal papers when a query asks for
>    [instance of] publications.
> 
>    I have some questions about this.
> 
>    First, what is the additional information that is being inferred about
>    journal papers? Is it that any instance of "journal paper" is
>    also an instance of "publication"? 

You are inferring that a particular instance of a "journal paper" is
also an instance of "publication." This may sound trivial, but without
the rule, you cannot make this conclusion.

> If so, isn't this information
>    something something that is not "new," but rather already present
>    in the rule "all journal articles are publications"? If not, what is
>    an example of the additional information being inferred about journal papers?
 
> 3. Your last sentence suggests to me that the search engine, in returning
>    "journal papers" when a query asks for "publications," is making some
>    kind of inference in order to provide these results.
> 
>    I have some questions about this as well.
> 
>    What inferences are being made, and how do these inferences relate
>    to the rule "all journal articles are publications"? Specifically,
>    do you mean by inference something different than the rule (implication?)
>    that "If X is a journal article, then X is a publication"?
>    The reason I ask is that I am confused about how, in the scenario you
>    lay out, inferences are different from rules (which to me seem to be
>    implications).

I was using the term rule in the same way that you are using implication
above. Typically, rules are logical implications, if something is true,
then something else must also be true. Inference is the process of using
rules to conclude something new. That is, if I have the fact "j1 is a
journal article" and the rule "If X is a journal article, then X is a
publication" then I can use the process of inference to deduce that "j1
is a publication." Sometimes, we may call this new fact an "inference."

----------------------------------------------------------------------------
> 
> On Thu, 21 Mar 2002, Jeff Heflin wrote:
> 
> > Jim,
> >
> > You raise some good points about that particular paragraph being
> > ambiguous. Let me try to rephrase it in a way that answers your
> > questions. If this rephrasing is sufficient, please let me know, so that
> > we can include it in the next version of the document. If it is
> > insufficient, let me know which parts are still unclear.
> >
> > This ontology can provide a terminology for describing content and
> > inferences sanctioned by the ontology can be used to improve the quality
> > of search on the portal. For example, an ontology might include
> > terminology such as "journal paper," "publication," "person," and
> > "organization." This ontology could include rules that state things such
> > as "all journal papers are publications" or "the authors of all
> > publications are people." These rules allow query systems to infer
> > additional information about journal papers and authors. Thus, a search
> > engine that is capable of inference can use the rules to return any
> > "journal papers" when a query asks for "publications" or return the
> > "author of a journal paper" when a query asks for "persons."
> >
> > Thanks for your comments.
> >
> > Jeff Heflin
> >
> >

Received on Monday, 25 March 2002 10:42:37 UTC