Introduction to Ontologies for GSA

1
Introduction to Ontologies for GSA

• What is an Ontology and how can it help?

2
What are Data Semantics and what is an Ontology?

• Ontology Definition
• An Ontology is a set of well-defined data
  semantics related to a particular domain that
  can be utilized for machine reasoning.
• Where XML can be used to specify the format of
  data (number of allowed characters, character
  sets, regular expressions, etc.) it does not
  inherently contain the meaning of the data.
• A pair of boots in system A may not mean the same
  exact thing as a pair of boots in system B.
  Ontologies solve this problem by defining
  meta-data classes codifying the meaning of data
  across an enterprise

• Semantics Definition
• In linguistics, the study of meanings In
  Computer Science, the formally defined meaning of
  data, distinct from syntax.
• Syntax is the format of the data
• Semantics is the MEANING of the data

W3C Semantic Web Stack
3
History of Semantics in Computer Science

• History of Semantics in Computer Science
• Mainframes and delimited files
• Early databases
• Modern databases
• SGML
• Xml and SOAP
• Ontologies

4
Ontology What is it?

• An Ontology is simply a way of expressing the
  semantic concepts present in a domain and their
  relationships to each other.
• As you can see from the image to the left, a
  basic concept such as a Person, can have 0-n
  attributes and 0-n linked concepts. These linked
  attributes and concepts are represented as
  classes, with internal attributes and
  relationships
• Ontological classes can be linked in a many-to-to
  many fashion, in multiple dimensions. There are
  no restrictions on inheritance or hierarchy.
• Actual data in the domain is associated with the
  ontology by associating the Object instance with
  an class in the ontology, effectively making the
  domain class instance a is-a instance of the
  concept. This does not preclude or impinge on any
  language based inheritance relationship as the
  RDF tag that links the domain instance to the
  ontology is an attribute of the domain instance.

5
Conceptual Relationships

• By linking the ontological concepts together, we
  can develop a mapping structure that will allow
  data to be collated automatically as it is
  discovered.

6
Ontology Instances
Below is a graph representing instances of the
person ontology shown in the last slide. The
graph below shows Bill and part of his family,
all of whom have traceable relationships to each
other and any other data available about them.
7
Adding Instances

• Next, we will take data already known, and using
  several available tools, apply RDF tags to the
  data based on the ontological concepts, creating
  instances of those concepts with associated data.
  An example of such instances for Terrorist
  Organizations is shown below

8
Vastly Improved Knowledge and Data Mining

• There is a query language for Ontologies, called
  SPARQL (Simple Protocol and RDF Query Language).
  This query language acts much like SQL, except
  its result sets consist of RDF tagged data
  pulled from 1-n heterogeneous data sources. Its
  like having access to all the data in an
  enterprise as a unified whole, without having to
  build and support a meta-data layer.
• In addition, because the relationships between
  the data elements are also represented as RDF
  tags, machine inferencing can be applied to
  discover relationships not readily apparent
  between data elements.
• A simple (although not precisely accurate)
  example of this is that if you had a DB with the
  triple flipper is-a dolphin and another DB
  with the triple dolphin is-a mammal a
  SPARQL query on flipper should return the
  information that flipper is both a dolphin and a
  mammal.

9
Graphical Querying ability

• There are some wonderful tools currently
  available for use with ontologies and SPARQL
  queries.
• To the left is a simple graphical SPARQL query
  that uses freely available web services to plot a
  location on a map by retrieving its latitude and
  longitude.

10
More Complex Examples

• As is seen on the left, multiple instances of RSS
  news feeds can be fed into a RDF tagging utility
  (in this case, the Thompson Reuters Calais
  engine) which will tag the RSS feeds with RDF
  tags allowing them to be processed and mapped.

11
Inferencing Graphs
As an example, if you wanted to trace the
relationships between Joe McGrath (the Former
Unisys CEO), the Unisys BOD, and other corporate
Boards, you could use an ontology that defined
those relationships and tag the appropriate data
with rdf tags on the fly, producing something
like the graph below.
12
Ontologies can be mapped to each other
Ontologies can be mapped to other ontologies,
that represent other data domains. This allows
heterogeneous data domains to inter-operate.
13
Ontology Transitioning Process
Legacy Applications, Databases, and Documentation
can all be used to create a core Ontology for the
Enterprise. Enterprise Modernization efforts can
be integrated with ontology design and
development.