View
216
Download
0
Embed Size (px)
Citation preview
Information Modeling: The process and the required competencies of its participants
Paul FrederiksTheo van der Weide
Paul Frederiks, Theo van der Weide 2
Position within Archimate
• The ArchiMate project is a research initiative that provides concepts and techniques to support an architect in the visualization, communication and analysis of integrated architectures.
• In this paper focus on: communication and analysis.
Paul Frederiks, Theo van der Weide 3
Requirements Engineering
• Discovering the purpose for which software is meant
• Identification stakeholders and their needs• Documentation stages:
– Analysis,
– Communication,
– Negotiation
– Decision making
– Subsequent implementation
• Closing gap informal - formal
Paul Frederiks, Theo van der Weide 4
Information modeling
• Identify involved information objects
• Resulting model used as base for communication and understanding
• Relying on common base for
understanding
• For example: (semi-)natural language
Paul Frederiks, Theo van der Weide 5
Motivation
Domain expert
System analyst
Paul Frederiks, Theo van der Weide 6
The information modeling process
Formalsemanticfunction
Informalsemanticfunction
Dialoguedocument
Paul Frederiks, Theo van der Weide 7
The goal
• Find a minimal (information) grammar capable to generate/accept the sentences of the informal specification
• Minimal in the sense that each formal concept is motivated form the informal specification.
Paul Frederiks, Theo van der Weide 8
Correct model
• Conceptual model as generative device
• Correctness:– Completeness principle: with respect
to Universe of Discourse– Falsification principle: with respect to
informal specification
Paul Frederiks, Theo van der Weide 9
Responsibilities
Domain expert:
Completeness
System analyst:
Falsification
Paul Frederiks, Theo van der Weide 10
Effectiveness
How well accomplish participants their share– How well can domain expert
• provide a domain description• validate paraphrased description
– How well can system analyst• map sentences onto modeling concepts• evaluate a validation
Number of cycles?
Paul Frederiks, Theo van der Weide 11
A theory for Information Modeling
• Our goal: try to find a theory for information modeling
• Main theorem for Information ModelingThe probability of a model being incorrect, as a function of the dialogue length, tends to zero for a combination of qualified domain expert and system analyst.
Paul Frederiks, Theo van der Weide 12
Refined
Paul Frederiks, Theo van der Weide 13
Refinement elicitation phase
• Collecting significant objects– D1: DE can provide complete set of
information objects– A1: SA can handle implicit knowledge
• Verbalization– D2: DE can provide any number of
describing sample sentences– A1: SA can handle implicit knowledge
Paul Frederiks, Theo van der Weide 14
Refinement elicitation phase
• Reformulation:– D3: DE can split into elementary
sentences
– D4: DE can reformulate in unifying format
– D5: DE can order sentences according dynamics in application domain
– A2: SA can validate sentences for
consistency
Paul Frederiks, Theo van der Weide 15
Refinement modeling phase
• Grammatical analysis and abstraction:– A3: SA can perform grammatical
analysis
– A4: SA can abstract sentence
structure, and match these
structures onto modeling
concepts
Paul Frederiks, Theo van der Weide 16
Refinement validation phase
• Production:– A5: SA can match abstract sentence
structure with concepts– A6: SA can generate new sample
sentences• Feed back:
– D6: DE can validate description– D7: DE can judge significance of
sample sentence
– A2: SA can validate sentencesfor consistency
Paul Frederiks, Theo van der Weide 17
Verification phase
• Verification:– A7: SA can think on an
abstract level
Paul Frederiks, Theo van der Weide 18
Summary
Paul Frederiks, Theo van der Weide 19
Conclusion
• Having these competencies at a sufficient level:– DE will eventually be complete– SA will guide DE in being complete
• Thus: information modeling will
lead eventually to a correct model
Paul Frederiks, Theo van der Weide 20
Base skills
Domain expert• D1: completeness• D2: describing• D3: splitting• D4: normalization• D5: ordering• D6: validation• D7: significance
System analyst• A1: implicit knowledge• A2: consistency• A3: grammatical
analysis• A4: modeling• A5: concretizing• A6: generation• A7: fundamental
Paul Frederiks, Theo van der Weide 21
Controlling natural language (1)
• Completeness:– D1: providing complete set of information objects– D2: providing any number of significant sample sentences– A1: handling implicit knowledge– A6: generating sample sentences
• Verbosity:– D3: splitting sentences– D4: reformulating in unifying format– D5: ordering sample sentences– D7: judging significance– A3: recognizing similarity– A4: abstracting sentence structures
Paul Frederiks, Theo van der Weide 22
Controlling natural language (2)
• Ambiguity:– D2: providing any number of significant sample sentences– D6: validating description application domain– A2: validating sample sentences for consistency– A6: generating sample sentences
• Consistency:– D2: providing any number of significant sample sentences– D6: validating description application domain– D7: judging significance– A2: validating sample sentences for consistency– A6: generating sample sentences
Paul Frederiks, Theo van der Weide 23
Controlling natural language (3)
• Mixed level of abstraction:– D6: validating description application domain– A3: recognizing similarity– A4: abstracting sentence structures– A5: matching natural language with modeling
concepts
Paul Frederiks, Theo van der Weide 24
Future research
• Introduction of open modeling concepts
• Extension of the dialog model
Paul Frederiks, Theo van der Weide 25
Open modeling concepts
• Natural language may be seen as a basis
• Other media might be more effective: a language with informal symbols and rules
• Solution: allow open modeling concepts.
• “Empowering a weak formalism by
negotiation”, in preparation
Paul Frederiks, Theo van der Weide 26
Extending the dialog
• In practice many stakeholders– particular view– goals
• The chatbox model– Dialog involves several participants– Sentence oriented– Subdialogs are possible
Paul Frederiks, Theo van der Weide 27
Thank you,
Questions?