Semantic Web

1
Semantic Web

• Linked Data

Scott E. Barasch barasch1
at umbc dot edu scottbarasch.com
2
What is Linked Data?

• Linked Data exemplifies the original vision of
  the Semantic Web as being a web of interconnected
  links of information such as those stored in
  FOAF, RDF, OWL or other files

3
Rules for Linked Data

• All data must be named with a URI
• This URI must be a valid URL
• There must be a page at this URL which contains
  the data that is represented by the URI Name
• This URL / URL should NEVER change
• Data should be interlinked between documents /
  files on the web

4
Challenges for Implementing Linked Data

• In the past, Semantic Web data was not published
  to the web
• It was stored in a zip file, and often stored on
  an external disk or tape media
• An example of this is an ontology which contains
  data about all of the Semantic Web researchers
• Recently this has changed, as the need for an
  interwoven mesh of linked data has become
  appearent

5
Goals for Linked Data

• Many different ontologies contain similar
  information for various data members
• I.e. Name, SSN, Birthday, Zip Code, Telephone
  Number
• These data members can be connected, to join the
  data from multiple ontologies into a giant
  collection of data, which can be commonly
  queried.
• The ultimate result would be to create an entire
  mesh web of all the ontologies in the world,
  where each ontology would be a node in a giant
  graph.
• That graph would be the Semantic Web

6
Linked Data Repositories

• DBpedia - a dataset containing extracted data
  from Wikipedia it contains about 2.18 million
  concepts described by 218 million triples,
  including abstracts in 11 different languages
• DBLP Bibliography - provides bibliographic
  information about scientific papers it contains
  about 800,000 articles, 400,000 authors, and
  approx. 15 million triples
• GeoNames provides RDF descriptions of more than
  6,500,000 geographical features worldwide.
• Revyu - a Review service consumes and publishes
  Linked Data, primarily from DBpedia.
• riese - serving statistical data about 500
  million Europeans (the first linked dataset
  deployed with XHTMLRDFa)
• UMBEL - a lightweight reference structure of
  20,000 subject concept classes and their
  relationships derived from OpenCyc, which can act
  as binding classes to external data also has
  links to 1.5 million named entities from DBpedia
  and YAGO
• Sensorpedia - A scientific initiative at Oak
  Ridge National Laboratory using a RESTful web
  architecture to link to sensor data and related
  sensing systems.

7
Problems with a Single Ontology

• Creating a single ontology out of all of the
  linked ontologies in the world would be a
  nightmare
• Data access and reasoning time would be
  astronomical
• The sheer load of a single user could possibly
  cripple the network
• No computer on the earth could realistically
  process and compute such a large amount of data

8
Alternatives

• Imagine a just in time access model for this
  single ontology
• The multiple ontologies would be linked by
  common data members (Name, Address, Zip Code, et.
  al.)
• Users or agents know ahead of time which
  ontologies they would wish to query
• These queries go only to the individual
  ontologies
• The data is returned to the user agent, which
  then parses the data, and connects the similar
  data members
• These data members are linked, and a local
  subset of the global single ontology is created
  for the extracted data

9
Interdependence

• Linking Data by itself is not enough
• We need to be able to follow those links, and
  combine ontologies so that we can combine the
  information stored in one ontology with the data
  stored in many other ontologies
• This merging of data allows us to gain more
  enhanced information, and sometimes can provide
  new information that is larger than the sum of
  all the information in all of the ontologies we
  are querying.

10
Examples of Linked Data Interdependance
11
In practice

• The concept of a data Mashup is how this is
  accomplished today
• A Mashup engine is the client side user agent.
• Web Services query Semantic Web data repositories
  and retrieve the requested data.
• The data is connected, and a greater meaning is
  discovered from small sets of disjoint data which
  are now connected.

12
What is a Mashup?

• A Mashup is a way of combining related data into
  a pictorial form using Socially Rich computing
  technology to make the data easy to read and
  understand
• Charts
• Graphs
• Websites
• Maps
• Tables
• Movies
• AJAX Rich Applications

13
Web 2.0 or Web 3.0?

• Web 2.0 is known as the Social / Collaborative
  Web
• Web 3.0 is another term used to express the
  Semantic Web
• The linked data is considered Web 3.0.
• The practice of pulling the data into the Mashup
  Engine is a mix between Web 3.0 and Web 2.0
• The practice of displaying the data in a Mashup
  is referred to as Web 2.0.

14
Examples of User Agents / Mashup Engines
http//www.jackbe.com/enterprise-mashup/
http//www-01.ibm.com/software/lotus/products/mash
ups/
15
Lotus Mashups Data Feeds
Data can be pulled from existing Enterprise
Datacenter Services, and also from feeds on the
internet or Semantic Web. Example input data can
include XML, RDF, LDAP, SQL, CSV, Office
Documents, RSS Feeds, Directory Servers, among
others.
16
Combining / Merging Data
Data mapping patterns, merging, looping, and
logical operations are all supported
17
Lotus Mashups - Wiring
18
Lotus Mashups A Mashup
19
Mashup Example
20
Mashup Example
21
Questions?