[sc34wg3] Semantic commitments as basis for Topic Map applications interoperability.

[Dmitry]

I tried to follow discussions about TAO/XTM/SAM and RM/TMM and these
discussions convinced me that question of semantic commitments is a key
issue for TM application interoperability.

TAO/XTM/SAM does following semantic commitments, I think:
. Topics as a way to represent subjects
. Relationships between subjects
. Collocation objective
. "instanceOf " as a special kind of relationships
. "subtype" as a special kind of relationships
. Resources as special kind of subjects
. Names as a special kind of relationships
. Occurrences as a special kind of relationships
. Associations as a way to represent "other" relationships.
. Context-based knowledge validity (scopes)


>From perspective of semantic commitments RM/TMM looks like this:
. Topics as a way to represent subjects
. Relationships between subjects
. Collocation objective

So, in fact, TAO/XTM/SAM -based topic map applications have more semantic
commitments. It means that if we try to map TAO/XTM/SAM -based topic map to
RM/TMM-based topic map we lose information and we cannot "restore" it back.

I suggest defining Topic Map Application based on semantic commitments. It
gives "simple" way to compare different models which look like "topic maps"
and also to discuss various TM "flavors".

This approach can provide some interesting results, I think.

>From "semantic commitments" perspective, Omnigator, for example, is not
really XTM/SAM application. It has more semantic commitments than XTM/SAM.
Using predefined PSIs Omnigator supports additional specialization of
occurrences.  When you do authoring for Omnigator you can take advantage of
these PSIs and end users will see information about subject in more
structured way.  And because Omnigator has more commitments than XTM/SAM,
any XTM/SAM-compatible application will be able to "understand" XTM/SAM
information in this topic map, just with fewer details.

Now, let's say I am not comfortable with TAO/XTM/ SAM definition of
"occurrences". I would like to extend it and represent additional attribute:
"occurrence strength".

So what can I do? I can define additional semantic commitment to existent
TAO/XTM/SAM commitments and define standard way to represent this extension.
I can reify each "domain" XTM/SAM occurrence as topic and specify additional
characteristic "occurrence strength".  I can "publish" this extension
somewhere and others can take advantage of this new feature. Software
developers can specify that their product implements TAO/XTM/SAM +
"occurrence with strength" commitments.

This semantic based approach can provide much better basic for TM
applications interoperability, I think.

Dmitry