EDUCE

Essential Domain + Usage Concept Extraction

Nik Boyd

educe, v.t.; to bring or draw out; to elicit; to evolve.

-- Webster's New Twentieth Century Dictionary - Unabridged - 2nd Edition

Motivations

• Exposes semantics directly from natural language
• Supports explicit selection from alternative conceptual models
• Builds commonality - the shared understanding of a problem and its solution
• Manages complexity through conceptual decomposition
• Serves as a natural precursor to object-oriented analysis and design
• Provides rigorous semantically based traceability

Are Methodologies Part of the Problem?

Software development methodologies

• typically begin only after a problem has been articulated
• transform problem descriptions into software solutions
• prescribe a relatively mechanical process from requirements to code
• tend to be solution oriented rather than problem oriented
• may prescribe how to collect usage requirements, but
• seldom prescribe how to describe a problem

In this regard, OO methods are no better than structured methods.

Development Requires Collaboration

People collaborate to specify and build software systems. See the associated software process model.

All participants need to share a common understanding, conceptual models of both
the problem domain and how software solves the domain problems.

Communication and Commonality are Critical

The importance of effective communication between participants cannot be overstated.
The quality of the communication between people determines their commonality

• commonality = the degree to which they share conceptual models of
  • the problem domain and
  • the usage requirements

Commonality ultimately determines the quality of the resulting software solution

• quality = the degree to which the software
  • solves the domain problems
  • satisfies the usage requirements

EDUCE Process Summary

The table below outlines the process steps and the results they produce. Note that the same process applies to both domain and usage analysis. See the associated analysis process diagrams. Note that some semantic explorations may enable certain kinds of syntactic normalization, and vice-versa. Thus, those two steps in the process may interact to create a useful conceptual model.

As an extensive example of this process, consider one of the problems used in the HP Fusion course - ECO Storage Depot. Also, you may find some background discussions on models interesting.

Domain/Usage Elicitation - Record discussions with domain experts or usage stakeholders

• Closely record source discussions
• Ask leading questions
• Elicit the vocabulary
• Generate and review the initial narrative

Syntactic Normalization - Simplify the syntax of domain/usage statements

• Use nuclear sentences
• Use singluar nouns
• Eliminate adjective prepositions
• Isolate verbs
• Use active verbs
• Use complete predicates

Semantic Exploration - Refine the model through semantic analysis

• Simplified language model
• Word class transformations
• Explore domain/usage terms
• Explore concepts: nouns + verbs
• Explore predicates: verbs + nouns
• Explore concepts: descriptive adjectives + nouns
• Explore quantities: limiting adjectives
• Explore qualities: adverbs + adjectives
• Explore quantities: articles
• Condense higher-order predicates into concepts

Semantic Organization - Find stable domain/usage concepts and predicates

• Identify stable concepts
• Construct initial conceptual models
• Construct a glossary

Model Validation - Verify models with feedback from domain experts and usage stakeholders

• Validate the conceptual models
• Assemble the vocabulary

Model Integration - Develop an integrated overview

• Integrate the conceptual models
• Regenerate the narrative