2/17/15
Ontologies and Master Data: Types and
Tokens, or Something Else?
After having laid out my argument that
ontologies and master data are not about “the same thing”,
Matthew West has repeated his belief that they are. But he has done
so without addressing my argument that they are not. I think the
point I tried to make is important enough and indeed that the case I
made for it (correct or incorrect) is solid enough that if Matthew
doesn't agree with me, he should say why he does not, and also
provide a cogent argument for saying that they are indeed about “the
same thing”.
Of course, “same thing” is a vague
enough predicate that for most comparisons of any but a trivial
nature, it will be possible to say that, in some sense, the compared
things are “the same”; but equally possible to say that, in some
sense, the compared things are different, i.e. not “the same
thing”.
I do not deny that many true things
have been said, in this thread, about ontologies and about master
data. If two things are not “the same thing” unless all
statements true of the one are also true of the other, then “same
thing” means “identity”, and by that standard, clearly,
ontology and master data are not “the same thing”.
But that is not the relevant standard.
I do not require Matthew, or anyone else claiming that ontologies and
master data are, in some sense, “the same thing”, to show that
they are the “identical, one-and-the-same thing”. Nor do I think
they mean that. However, I do ask that they do more than make some
statements that are true of one, but not of the other. The
distinction, or the absence of one, must be based on a more
substantive issue than that.
For me, I use a theoretically
well-grounded distinction to support my contention, and I explain in
unambiguous terms how that distinction applies to master data and to
ontologies. If Matthew or anyone else disagrees with me, I would
appreciate hearing what is wrong with my argument.
To adumbrate that argument from some
messages of a few days ago, I said that ontologies are about types
and master data is about instances/tokens or types. Now since the
type/token distinction is as clear as the distinction between sets
and their members, or sentences and their utterances/inscriptions,
and is a staple of formal logic, I claim that the distinction is
theoretically well-grounded. It is also well-established in
introductions to logic. Indeed, it is the basis of the distinction
between first-order and higher-order logics, first-order logic being
about instances and higher-order (second-order) logic required to
formalize types of those instances, including types of kinds of
things, types of properties of things, and types of relationships
among things.
Given all this, the next question is:
does the distinction apply in this case? Are ontologies about types,
and not about tokens? From a theoretical point of view, going back to
Aristotle, they clearly are. From the point of view of current usage
of the term “ontology”, I have never found anything called an
ontology that was not about types, and was about tokens.
And is master data about tokens, not
about types? I think it is clear that it is. For example, from the
Wikipedia entry for “master data”, we have:
“Material Master Data is a specific
data set holding structured information about spare parts, raw
materials and products within Enterprise Resource Planning (ERP)
software. The data is held centrally and used across organizations.
….. Vendor Master refers to the centralized location of information
pertinent to the Vendor. Often this will include the Legal entity
name, Tax identification and contact information.
Clearly, this quotation is talking
about spare parts, not Spare Part as a type, nor to types of Spare
Parts, to specific instances of raw material, not to a type labeled
“Raw Material”, to date about specific vendors, not about Vendor
as a type, and so on.
And I have another argument showing
that the type/token (or type/instance as it is sometimes called)
distinction does apply to ontologies and master data. As I explained
in the earlier messages, relational tables represent types of things,
and their columns represent types of properties of those things or
relationships among those things. (The fuller account can be found in
Chapters 4-6 of “Bitemporal Data: Theory and Practice” (BDTP).
Yet SQL does not manages and retrieves data about types of things. It
manages and retrieves data about instances of those types, and rows
in relational tables are each descriptive of one instance of the type
represented by the table it is in.
I think that what I have said is true,
and also that it is important (although I haven't said anything about
that second point so far). If Matthew or anyone else thinks that what
I have said is not true, I would appreciate hearing why they think
that, and specifically what about my argument is wrong.
I do not claim that any reading about
ontologies and master data, no matter how extensive, would make what
I have said obvious. The situation is at best, I think, a matter of
the trees hiding the forest. It seems to me, with all due respect,
that Matthew and others contributing to this thread have pointed out
various trees, and in some cases tried to describe the clump those
trees comes from (i.e. what they have in common, that commonality
being the proffered distinction/similarity between ontologies and
master data).
But perhaps the problem is that I am
wandering in the wrong forest.
I'd like to know what others here
think.
With my thanks in advance,
Tom Johnston
In my experience, the distinction is :
MDM is a discipline concerned with managing data models and key data values and codes. For example, some people
consider the list of customer contact information as master data, the names and columns of RDBMS tables as master data, country code lists as master data, etc. So, MDM apps are built using data models whose data values are about other data models/values/codes used in their enterprise. MDM has a fuzzy definition, as noted by others, but most people I’ve worked with mean what I’ve suggested when using the term.
Ontologies are data models written using logic-based languages and can cover whatever scope is of interest. So, you can have an ontology whose scope is supporting MDM, and your MDM app may contain names of classes and properties in operational ontologies. Some people require that data models be about the real world to be considered an ontology, but others are less worried about that criteria when using that term and really only care about the data
model being written using logic-based languages.
So, simply put MDM apps are about managing other apps in my enterprise, and ontologies are a kind of data model that can be the basis for any enterprise app.
Cheers,
David
UK +44 7788 561308
US +1 336 283 0606
Dear Erick,
This is largely a question of the different language used by different communities. People who work with RDBMS and are struggling to integrate data across different databases will talk about Master and Reference Data, and use tools like SQL and ETL. People who use OWL or RDF and Prolog are likely to talk about ontologies, and be looking at performing reasoning across their ontologies. What the ontologies and master data represent is essentially the same thing.
Regards
Matthew West
Information
Junction
Mobile: +44 750 3385279
Skype: dr.matthew.west
This email originates from Information Junction Ltd. Registered in England and Wales No. 6632177.
Registered office: 8 Ennismore Close, Letchworth Garden City, Hertfordshire, SG6 2SU.
Hi,
I need some help to better define the line (sometimes apparently grey) between master data and ontologies.
We all, at least in this forum, know that there are
several definitions for both terms.
I guess most of us are familiar with Gruber's one: a formal specification of a shared conceptualization.
In the case of master data:
- 'entities, relationships, and attributes that are critical for an enterprise and foundational to a key business process and application systems'
- 'is the consistent and uniform set of identifiers and extended attributes that describes the core entities of an enterprise'.
What are the key components to differenciate master data and ontologies?
What is common to both artefacts?
From what I have seen, sometimes the border between them seems indeed relatively
grey... which seems to be the product of having ontologies as glue components of disparate master data. Also, there seems to be a continuum between them (as in the databases and knowledge base thread in this forum). Anyway, I would appraciate reading your thoughts about it.
_________________________________________________________________Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/ Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/ Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxxShared Files: http://ontolog.cim3.net/file/Community Wiki: http://ontolog.cim3.net/wiki/ To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J