Friedland 20040428

Issues in Negotiating Multiple Semantic Models : Issues in Negotiating Multiple Semantic Models LeeEllen Friedland Center for Integrated Intelligence Systems The MITRE Corporation lfriedland@mitre.org April 28, 2004


The Semantic Web : The Semantic Web Vision To provide 'a common framework that allows data to be shared and reused across application, enterprise, and community boundaries.' (W3C) Semantic models are key Different types of models Classification systems, taxonomies, thesauri, topic maps, ontologies, etc. Different ways to represent these models in machine-usable form Formally expressed Standards-conformant


Semantic Models : Semantic Models Where do semantic models come from?


More Questions About Semantic Models : More Questions About Semantic Models Are they already being used in this (mostly) pre-Semantic Web world? How do you create models in areas where there are none? How do you decide what type of models to create for different purposes? How do you know if you have the models you need? How do you judge whether a model is correct? How can you be sure a model will work with a specific application or serve a certain goal? Are there different requirements for models to be used in different contexts? What do you do if you have legacy models that need to be updated? What do you do when you have legacy models and you also have to create new ones?


Key Issues : Key Issues Understanding semantic information models How semantic information models are used in context Strategies for negotiating environments in which multiple heterogeneous models coexist Understanding these issues provides a foundation for Semantic Web applications


The Importance of Models : The Importance of Models Semantic Web is new Semantic models are not Semantic models are used throughout everyday life Semantic models are used throughout work activities Semantic models are not about IT applications They represent the way people think about the things they know and how they make sense of the universe Why does this matter?


People versus IT? : People versus IT? Framework for working with Semantic Models #1 - Cultural and social contexts Concepts and relationships must reflect real-world values and beliefs Authenticity, fidelity #2 - Content-centric Models must reflect the expert knowledge and judgment of specialists Domain expertise


Two Perspectives Are Needed : Two Perspectives Are Needed Information management / content management Content inventory, user requirements, lifecycle management requirements, metadata design, controlled vocabularies, semantic models (classification systems, taxonomies, thesauri, topic maps, ontologies, etc.) Cultural / social (ethnographic approach) Values, attitudes, beliefs, customs, work culture, work practice, social interaction, etc. Background in IT area is helpful, but often not essential Provides context for some future use Semantic Models should be created and managed over time independent of IT implementation(s)


Semantic Models at the IRS : Semantic Models at the IRS IRS Electronic Tax Administration Public Portal Branch IRS.gov Content management system eGovernment initiatives E-Government Act of 2002 (H.R. 2458/S. 803) Executive owner Partner MITRE developed a Metadata Strategy and Plan (Spring 2003) Based on domain metadata meetings with 15 business units, over 60 staff members IRS.gov subject taxonomy and thesaurus Enhancement and upgrading of content management program *The views expressed here are those of the author and are not necessarily in accordance with the views of the Government.


The Semantic Landscape : The Semantic Landscape A variety of Semantic Models in use, for example: Legacy taxonomies Uniform Issue List (Office of Chief Counsel) 'Index Entry Terms' (Forms and Publications) Internal Revenue Manual Index (SPDER) Taxonomies proposed or in development eLearning eFOIA Topic Map pilot (Multimedia / ETLA)


The Hidden Semantic Landscape : The Hidden Semantic Landscape Hidden in plain sight Semantic models in widespread use No one mentioned them They had no name They had no owner Not documented Ontologies Tax Administration Tax Law 'Master' vocabulary


The Semantic Landscape: What was Expected : The Semantic Landscape: What was Expected Domain Knowledge (Community Level) Domain Knowledge (Community Level) Domain Knowledge (Community Level) Domain Knowledge (Community Level) Domain Knowledge (Community Level) Domain Knowledge (Community Level) Domain Knowledge (Community Level) Domain Knowledge (Community Level)


The Semantic Landscape: What was Found : The Semantic Landscape: What was Found Domain Knowledge (Community Level) Domain Knowledge (Community Level) Domain Knowledge (Community Level) Domain Knowledge (Community Level) Domain Knowledge (Community Level) Domain Knowledge (Community Level) Domain Knowledge (Community Level) Domain Knowledge (Community Level) Tax Administration Ontology Tax Law Ontology


What is the Enterprise? : What is the Enterprise? The concept of the community- or domain-level is clear The concept of the enterprise-level is less clear and can change according to context Are the terms 'enterprise' and 'community' adequate to express variable relationships between different levels of organizations, domains, and ontologies, and the different contexts in which Semantic Models are used? Tax Administration Ontology Tax Law Ontology


Key Characteristics Of Semantic Models in Context : Key Characteristics Of Semantic Models in Context Implicit versus explicit Implicit Semantic Models can be well-developed and widely-used Informal versus formal versus formally represented When informal Semantic Models are being used successfully they, or their applications, will usually benefit from higher levels of formalization The need to formally represent a Semantic Model is tied to the need or wish to use it in an IT context Non-automated versus semi-automated versus optimally automated The degree to which a Semantic Model is automated in a technology application has no bearing on its fidelity to the concepts it represents


Characteristics of IRS Semantic Models : Characteristics of IRS Semantic Models


Conclusions : Conclusions The future success of the Semantic Web is dependent on the integrity of the Semantic Models we use, represent, and implement Complex organizations will almost certainly have legacy Semantic Models in use with varying profiles of key characteristics Understanding the Semantic Landscape requires at least two perspectives: information management and cultural/social Semantic Models represent concepts created and used by humans The employment or adaptation of these models in a technology application is only one of many possible uses and therefore should be designed and managed as a distinct activity