Re: SUO: Re: SUO Ballot with 2 Questions
John,
Thanks for your comments. I appreciate your desire to ensure that we achieve something of real utility from this process rather than a coronation of a single but inadequate ontology. I think you're addressing a real danger of which we need to be quite wary. However, I am not convinced that proposal #2 is the way to avoid it. More comments below.
best,
Mike
At 10:36 PM 6/2/2003 -0400, John F. Sowa wrote:
>Eric and Mike,
>
>I share your wishes and hopes. It would be
>wonderful if there were a single perfect
>(or at least very, very good) ontology that
>everybody could agree was an outstanding
>achievement worthy of universal acclamation.
>
>Unfortunately, nothing like that exists. And
>of the two content ontologies under consideration,
>neither group of developers is willing to concede
>to the other. Furthermore, there are many other
>potential candidates that are not represented.
>See John Bateman's ontology portal:
>
>
>http://www.fb10.uni-bremen.de/anglistik/langpro/webspace/jb/info-pages/ontology/ontology-root.htm
>
>John lists over 20 general ontologies, all of which
>could be considered either competitors or potential
>supplements to SUMO or OpenCyc. Many of the developers
>of those ontologies, such as Nicola Guarino, for example,
>have very serious complaints about both SUMO and OpenCyc.
>
>I cannot imagine Nicola or others on Bateman's list
>agreeing to endorse any ontology that does not allow
>them an opportunity to contribute their ideas (which
>motion #2 does). Unless we can provide a framework
>that allows them to contribute modules to a larger
>effort, they will go their own way. And they have
>enough expertise to develop a serious independent
>standard of their own -- which they are actively doing.
But there is an easier solution here, isn't there? Guarino et al can submit their own work as possible standards or they can offer extensions and revisions to SUMO and/or OpenCyc or what have you. I think that the DOLCE ontology, or a future mature version, is worthy of consideration. Failing that, we could and should discuss the sorts of objection that some raise against the candidate standard ontologies and determine whether they're pertinent, and if so whether they can be overcome. If there are problems with the candidates, this is an argument to change or reject them, not an argument to create an environment within which any number of sub-par ontologies can peacefully coexist.
>Motion #2 is, I believe, a good compromise for working
>toward a consensus. It doesn't guarantee that an ideal
>standard ontology will emerge, but without something
>like motion #2, the SUO group will fall apart --
>thereby shutting down this effort altogether.
To my mind motion 2 generates a system within which our putative standard is nothing more than a place into which any Tom, Dick, Harry (or, if we're lucky, Adam or Doug) can insert an ontology and the result is a standard which is only of slightly more use than the Bateman URL you give above (i.e., not useless at all, but not really a standard) or we're simply delaying the kinds of decisions that you claim will shut down this effort altogether. Either way, we're not going to generate a real SUO.
I'm a bit more optimistic. I believe that having identified candidate starter documents we can commence work on criticizing and improving them with an eye to eventually selecting one that is of use as a real standard. If the critics of SUMO and OpenCyc are willing to throw in their two cents worth, and indeed many have, so much the better. If others want to proffer candidate ontologies as starter documents, they should do so and we can consider them on their merits qua SUO. Of course, there may be some who choose not to participate in the process and then after all is said and done will complain that the SUO is useless or faulty but to my mind it's hard to give a good argument that we need to pander to this kind of non-participant and weaken the standard in such a way so as to be ultimately acceptable to any and all, even those whose objections we do not know.
>The rules allow voters to change their votes before
>the closing deadline. I address your concerns in
>more detail below, and I hope that you will reconsider
>your votes.
I'm happy to hear this, perhaps you'll change yours? ;>)
<snip>
>> Responding to a mandate to generate a SUO with a general
>> registry containing distinct but mostly overlapping upper
>> ontologies seems a bit like returning two taped together
>> dictionaries and specifications on the kind of tape used
>> in response to a request to generate a standard dictionary
>> for the English language (a standard to which the desirability
>> of extensibility and modularity would also apply).
>
>The process of building any dictionary begins with analyzing
>and organizing all the available resources -- such as a large
>corpus of word usage together with masses of reference works
>(e.g., earlier dictionaries and encyclopedias) and large
>amounts of researh into particular issues of lexicography
>and linguistics.
Indeed, but the standard itself should not be the compilation of materials. I worry that this is what the "registry as standard" proposal naturally leads to.
>That is what motion #2 proposes that we do.
>
>> FWIW, I encourage the OpenCyc group to resubmit OpenCyc as a stand
>> alone ontology. I would readily vote for it as a good starting point
>> for an upper ontology. Should it and the SUMO effort both be
>> accepted as starter documents, each could be developed and the
>> group could discuss and decide at the end of the day which is the
>> most deserving of IEEE standard status.
>
>I encouraged the OpenCyc group to submit a separate motion, but
>they were comfortable with the joint motion.
>
>A shoot-out between independent SUMO and OpenCyc projects
>would be would be worse than useless. It would do nothing
>to encourage either group to collaborate with the other,
>nor would it make either ontology more attractive in comparison
>to the other dozen or more ontologies on Bateman's list.
>
This isn't obvious to me. Suppose that we have ontology experts, people like you and Nicola Guarino, offer critiques of the candidates. The developers could choose to respond accordingly, with arguments as to why the criticisms aren't salient, with alterations or by simply ignoring the critiques. At the end of the day we'll have ontologies that have benefited from the input of people on the list and the refinement efforts or we'll have upper ontology standards candidates that have been shown to be flawed or inadequate and we can give them the thumbs down for principled reasons after suitable discussion.
The key here is that the respective candidates improve and take seriously the suggestions and criticisms of the people in this group. I don't think that collaboration between them will be necessary for either (any) of them to generate an attractive SUO candidate. However, suppose that it is. Suppose that good arguments are made asserting that the ultimate standard will require the integration of components from both candidates. Again, the developers of SUMO and OpenCyc (and any other proffered ontology) can choose to reject these arguments and we can choose to ultimately reject their respective candidates or they can choose to collaborate and generate a SUO candidate that has a chance of being adopted by the group. Where is the risk here?
___________________________________________
Mike Pool
Information Extraction & Transport, Inc.
1911 North Fort Myer Drive, Suite 600
Rosslyn, VA 22209 USA
(703) 841-3500 x632
(703) 841-3501 Fax
mpool@iet.com