notes in attachment - Confluence

engineerbeetsΤεχνίτη Νοημοσύνη και Ρομποτική

15 Νοε 2013 (πριν από 3 χρόνια και 6 μήνες)

453 εμφανίσεις

12/20/2012


R3 semantics phone call


Present: Michael, John, Maurice, Karen


Action list:

1.

Karen

will distribute minutes of mtg. DONE

2.

Karen

will develop draft list of Ontology resource attributes, based off MMI
ORR. Post off CIAD APP Resource
Model

confluence page. Target: 12/21

3.

Karen

will develop draft list of Attribute resource attributes (say that 3
times fast), based off existing Filemaker content. Post off CIAD APP Resource
Model
. Target: 12/21

4.

Karen and John

will initiate draft list of Assoc
iation resource attributes.
target: LCO?

5.

Karen

will develop a list of existing resource attributes to expose in the ION
Resource Ontology

6.

John

will provide information on the inferencing engine and related tools
used by MMI ORR. Target: ??

7.

Michael/Maurice

will add content to the “Semantic page on arch” (Question:
is this
https://confluence.oceanobservatories.org/display/syseng/CIAD+DM+OV+V
ocabulary+Management
?). Based on notes from today, plus perhaps diagram
and further detail. Target: LCO


Note: Action
s 2 and 3 together will complete Karen’s Jira task 627.


Semantic Needs



register vocabs/ontologies



use them in services



represent them in UI


Will add 3 new resource types to describe semantic assets:


1.

Ontology



MMI ORR provides an initial set of
attributes for consideration



can be a pointer to information in external repository, or internal
representation (e.g. RDF or similar)

2.

Attribute definition


needs to define attribute’s meaning, and/or also the
value set.



Initial content will be derived f
rom the fields in the current Attribute
Filemaker Database



the attribute concept may be supported by an ontology



the allowable content for an attribute may also be supported by an
ontology (So CV could be an attribute “type”)

3.

A
ssociation definition


are
both of the above are concepts



sim
ilar

to above



needs to define meaning,
as well as constraints (one to one, one to
many,
etc.)



look at existing systems for models (sim to RDF,e.g.)



System functions related to semantics



search


i.e. pull back all “temp
erature” records even if have different param
names.



query


similar to above



navigation


the system gives the user a set of alternatives. I.e. do you want
to drill down/limit by certain types of items. Needs knowledge that items
participate in same ca
tegory. Comes out of ontologies and inference, but can
also be authored. R2 resources such as Catalog and View not implemented
this way yet, but structured in this direction.



inference


low level function that helps search/nav to work



transformation


e.g. for importing external data that has a different known
vocabulary into our terminology. Or unit conversion. Transform of resources
based on ontology. Implicit or explicit.



externalization


e.g. presenting our metadata in standard formats such as
IS
O. Can be relabeling names, or converting units, etc.



identification of versioning (at uri/concept level)


inspiration is MMI. Latest
version, current version, history, etc. Must be able to go back to identity of
that concept at the time applied. Not c
lear how back end would support this


current resource framework does not hold history.

o

versioning applies to the definition level, and the content level, of a
resource. Is it imp to know prior values of the resource? Def need to
know prior definition of

the definition. (Could push versioning into
the ontology representation world…but…)


Additional discussion:



Will need to create an “ION Resource Ontology” based on our attribute terms.
And later may decide to map this to external ontologies. This can be
extracted from the YML attributes and associations and converted to OWL.
Karen

will identify the first set
of terms to expose, based on how user
-
facing
they are (at present will focus on most user
-
facing attributes).



We will want to perform inferences, and cache them for faster access. But
that cache is likely not a first class resource.



Linked data: there are

considerations regarding how far OOI wants to go
with respect to exposing content via Linked Data.



Parameters: consider parallels. Parameters are like attributes. Parameter def
is like attribute def. Parameter dict is similar to object type.



May need t
o include:

o

an inference engine.

o

probably not a repository


would do as interface.

o

May need analytic store (i.e. to store an external ontology).

o

May also consider a semantically
-
enabled data store (BigData?...tools
exist that may help).

o

Note that one

existing tech may serve all 3 needs

MMI has a solution;
John

will provide further info.