Skip to main content

Notes from Thursday Semantic Technology Conference

Posted by haroldcarr on June 18, 2009 at 3:11 PM PDT

Here are my notes from Thursday at

www.semantic-conference.com

==============================================================================
Thursday, 6/18/2009

Conference Attendance: 1170

------------------------------------------------------------------------------
Keynote General Session 8:30AM - 09:30AM

Semantics at The New York Times
Larson, Robert, New York Times Company
Sandhaus, Evan, New York Times Company
Conners, Christine, moderator
Business / Non-Technical
Publishing

What the Times means by Semantics
- Any tool that discovers/leverages metadata to
  make content more accessible

The Morgue
- File cabinets of clippings, photos,... of Times and other papers
- Reporter seeking info on story would go to morgue and search

The New York TImes Index
- Where info can be found in printed version

NYTimex.com
- MS Word   -> Newspaper CMS -> tagging software -> web cms
  reporters    editors           section, pagen
  links         byline            newsdesk
                type
                biographic
- Topic pages
  Aggregate from NYT and other locations

Future
- Improve workflow
- Digitize 5 million photos from morgue
- tag NYT blogs
- Tag user content
- Allow users to tag

Times APIs
- Times Tag
- Article Search API

New York Times Annotated Corpus
- DVD of every article+metadata 1987-2007
- corpus.nytimes.com
- non-commerical: $300

------------------------------------------------------------------------------
Conference Sessions 9:45AM - 10:45AM

Visualizing RDF
Franz - Gruff

Kieth Sutton - silvafug.org - Silicon Valley Flex User Group

www.diagramic.com
EX: www.forbes20.com, www.diagramic.com/v/?p=google
FlexParts
(Adobe AIR application)
- Visualizing Freebase data

SEED FOR NEXT YEAR'S CONFERENCE:
- Visualizing RDF with JavaFX

------------------------------------------------------------------------------
Conference Sessions 11:00AM - 12:00PM

--------------------------------------------------
SQUIN
squin.org
query the web of linked data
Uses SINDICE
EX usage: turn2live.com

--------------------------------------------------
Practical OWL 2 RL Reasoning Via Fast Forward Chaining Inference Engines
Matheus, Christopher John, VIStology, Inc.
Technology - Intermediate
Foundational Topics

OWL 2 RL
- Implemented by rule-based forward chaining inference engines
- Implemented as collection of 75+ implication and consistency rules

Unsupported/Restrictions
- owl:cardinality, owl:minCardinality
- owl:qualifiedCardinality owl:qualifiedMinCardinality
- owl:DisjointUnion
- owl:ReflexiveObjectProperty
- owl:topObjectProperty, owl:bottomDatatypeProperty
- owl:topDataProperty, owl:bottomDataProperty
- xsd:real, wsd:rational
- URI cannot be both a class and an individual
- owl:maxCardinality restricted to 0 & 1
- Data ranges restricted to sxd datatypes and intersection of them
- Axioms designed to avoid inferring existence of individuals
  not explicitly mention in knowledge base
- Does not include axiomatic triples (but impls may include them)
  (e.g., characteristics of elements of the language)

Types of inferencing
- Entailment
- Classification (entailment of rdf:type facts)
- Consistency checking

P rdfs:domain C
I P V
-----
I rdf:type C
(just by giving I property P then you can infer its type)


B rdfs:subClassOf A
I rdf:type B
------------
I rdf:type A


parentOf owl:inverseOf childOf
John parentOf Diana
------------------
Diana childOf John


C owl:hasValue V
C owl:onProperty P
I rdf:type C
------------
I P V


C owl:hasValue V
C owl:onProperty P
I rdf:type C
------------
I rdf:type C

C1 owl:disjointWith C2
I rdf:type C1
I rdf:type C2
--------------
INCONSISTENCY

--------------------------------------------------
Ideas for Future Semantic Applications
Sarris, Tony, Unisys Corporation - Systems & Technology
Sweeney, Peter, Primal Fusion
Business / Non-Technical
Business & Marketplace

PrimalFusion

------------------------------------------------------------------------------
Product Seminars/Field Trips 1:00PM - 5:00PM

--------------------------------------------------
A Programmer's Introduction to Pellet: How to Build Ontology-based Semantic Applications
Sirin, Evren, Clark & Parsia LLC
Smith, Michael A, Clark & Parsia LLC
Technology - Advanced
Semantic Applications

clarkparsia.com/pellet/tutorial

What is Pellet?
- OWL-DL reasoner
- Supports nearly all of OWL 1 and OWL 2
- Sound and complete reasoner
- Written in Java and available from  http://clarkparsia.com/pellet

Talk Coverage
- Introduction to reasoning with Pellet
  Basic reasoning concepts
  Using Pellet from command-line
- Programming with Pellet
  Available APIs and usage
  Reasoning and query answering
  Explaining inferences
- Advanced Pellet features
  Closed-world data validation with Integrity Constraints

Running Example: POPS
- Expertise location in a large organization
  Based on POPS application in NASA
  Multiple sources containing personnel data: contact
   information, work history, evidence of skills,
   publications, etc.
  Find people that satisfy certain conditions
- Several challenges
  Integrate data from multiple sources
  Ensure data consistency
  Query with inferencing
  User interface (not covered in this talk; see jSpace)

Demo of JSpace/POPS

Building the Example
- Author ontology schemas
  Validate and debug schema definitions
- Connect multiple schemas
  Simple ontology alignment
- Validating instance data
  Identify and resolve inconsistencies in the data
  Closed world data validation with Pellet Integrity
  Constraints
- Reasoning with instance data
  Answer queries over combined data using Pellet
  Scalability and performance considerations

OWL in 3 Slides (1) - ENTITIES
- Class: Person, Organization, Project, Skill, ...
- Datatype: string, integer, date, ...

- Individual: Evren, C&P, POPS, ...
- Literal: "Evren Sirin", 5, 5/26/2008, ...

- Object Property: worksAt, hasSkill, ...
- Data property: name, proficiencyLevel, ...

OWL in 3 Slides (2) - EXPRESSIONS
- Class expressions
  and, or, not
  some, only, min, max, exactly, value, Self
  { ... }

- Datatype definitions
  and, or, not
  <, <=, >, >=
  { ... } 

OWL in 3 Slides (3) - AXIOMS
- Class axioms
  subClassOf, equivalentTo, disjointWith

- Property axioms
  subPropertyOf, equivalentTo, inverseOf,
  disjointWith, subPropertyChain, domain, range

- Property characteristics
  Functional, InverseFunctional, Transitive,
  Symmetric, Asymmetric, Reflexive, Irreflexive

- Individual assertions
  Class assertion, property assertion, sameAs,
  differentFrom

Reasoning in OWL
1. Check the consistency of a set of axioms
   Verify the input axioms do not contain contradictions
   Mandatory first step before any other reasoning service
   Fix the inconsistency before reasoning
     Any consequence can be inferred from inconsistency
2. Infer new axioms from a set of axioms
   Truth of an axiom is logically proven from asserted axioms
   Infinitely many inferences for any non-empty ontology
   Inferences can be computed as a batch process or as
     required by queries

Classification
- build tree of subClass relationships

Realization
- assign individuals to their most specific class membership

Protege 4 does not come with Pellet built in
- A plugin is available
- Use protege "check for plugins"