Getting Started with Semantics gin the Enterprise

Bradley ShoebottomBradley ShoebottomNovember 10, 2010, AWOSS, Moncton, NB

Introduction

Should your enterprises’ first ontology look like this (and take 2 years to get there)?like this (and take 2 years to get there)?

A representation of the internet(Credit: Bill Cheswick Lumeta Corp )

2

(Credit: Bill Cheswick, Lumeta Corp.)

Contents

Getting StartedSelecting Semantic ProjectSelecting Semantic ProjectCommon Semantic ProjectsOntology ModelingOntology ModelingOntology Software ToolsApplication developmentApplication developmentFinding Semantic SpecialistsP j t M tProject Management

3

Getting Started

All new efforts, including semantics require:• BudgetBudget

− Formal (project or RnD effort)− Informal

S k h ld • Stakeholder support− Listen to issues− Brief them about project

Usersp j

− Keep them up to date− Provide post project report

ProjectProject

MgmtIT

4

Selecting Semantic Project

To help ensure success and build support for semantics, select a project that:semantics, select a project that:

• Starts with a simple problem (KISS principle)• Will impact a large number of employees• Solves a real problem• Has a user base • Has a business case comparison

− Do nothing cost− Traditional IT solution costTraditional IT solution cost− Semantic cost

• Know when to draw the line

5

Common Semantic Projects

Some ideas for a first semantic project:• Semantic integration of various data bases *Semantic integration of various data bases• Knowledge Exploration *• Semantic annotation• Automatic document tagging• Auto-summarization of a document• Business intelligence mash-ups• Decision making support

Common metadata vocabulary for self annotation • Common metadata vocabulary for self annotation of documents

6

Ontology Modeling

There are a number of ontology modelling considerations:considerations:

• Approaches to modelling• Semantic precision of the model• Reuse (Modularization)

7

Ontology Modelling: Approaches

• Bottom-up

• Top-down

• Mixed Methodology( k b t)(works best)

8

Ontology Modelling: Semantic PrecisionH h S i i i i d d?How much Semantic precision is needed?

• Do not model everything initially

9

Telecommunications Hardware Ontology Excerpt

Ontology Modelling: Semantic Language SelectionD d d i d l /l i i iDepends on desired rules/logic expressivity

10

Ontology Modelling: Reuse

Design ontologies to be reusable• As open source or within enterpriseAs open source or within enterprise• Keeps costs down

− For next project− For maintenance (collaborative/distributive)

Image Credit: Obitko.com

11

Ontology Engineering Methodologies

Ontology engineering methodologies vary by:• ComprehensivenessComprehensiveness• Conceptual vs.. implementation• Decentralized vs.. centralized ontology

development• Developmental vs.. Support processes

I i f d h d lInnovatia preferred methodology:• Melting Point by Alex Garcia

T k it ti hTake an iterative approach:• Develop one part of ontology, test, correct, test

Repeat with next part

12

• Repeat with next part

Ontology Software Tools

Wide variety of new tools needed• Ontology Editors • Rules engine/reasonerOntology Editors • Text mining

g• End user interface• Query languages

Best Practice:• Learn tool before project, or plan time in project

for learningConsider open source versus commercial• Consider open source versus commercial− May get you going faster

13

Data Triple Stores

Triple stores store triple assertions for faster querying of data:querying of data:

• Open source versus commercial?• Consider an open source triple stores to start with

until you understand your requirements for:− Scalability (number of triples) − Uploading speedUploading speed− Inferencing speed− Querying speeds

h h h k f • Use Lehigh University Benchmark (LUBM) for benchmarking

14

Application development

Applications are needed to present data to end users. end users. Best practices:

• Define user queries up frontq p• Initial interface does not have to be fancy• Plan for iterative usability/feature

implementation

15

Maintenance of a semantic effort

Semantic projects require ongoing support no matter the development approachno matter the development approach

• Ontology modification− Centralized (top-down)− Decentralized contribution(bottom-up)

• Analysis of search requests to publish new search Analysis of search requests to publish new search patterns (FAQ or suggested search)

• Business processes needed for maintenance support− Mix of centralized for foundational ontology and

decentralized for changing parts of ontology

16

Finding Semantic Specialists (1 of 2)

Staff you may need:

New Role Description

Y Scientific Advisor -Solutions Architect

Guides semantic effort

N Domain Expert Knows content or data very well

Y Ontologist Creates ontology

Y Text Mining Engineer Configures text mining processing softwareY Text Mining Engineer Configures text mining processing software

Y Logics Reasoner Develop axioms for business rules

N Applications Developer Creates end user interfaces

N D t b A hit t U d t d i ti DBN Database Architect Understands existing DB

N Information Architect Determine user needs and user testing

N Business Analyst Define requirements, build business case

17

N Project Manager Keeps the rest of the team on track!

Finding Semantic Specialists (2 of 2)

Where can you find the staff?• Internally: may require some trainingInternally: may require some training• Local University

− Professors (high level advice)− Post-graduate− Students

• Job market (contract): roles/deliverables will Job market (contract): roles/deliverables will need to be clear for best results

18

Project Management

Semantic projects can conform to normal project managementproject management

• Give extra time for learning/experimentation• Iterative testing based on small component

development (Agile-like)• Can have parallel development

B t t t t k d b d t f • But, expect to encounter unknowns and budget for it

• Get a Project Manager!Get a Project Manager!

19

Useful Resources

• Obitko.com• Semantic Web (Getting ( g

Started)• Semantic Exchange• Semantic Web for the Working

Ontologist (book)

20

Conclusion

Emotion Ontology

21

Emotion Ontology