Stream Forge – Stories - mdebellis/SemanticKG-Design GitHub Wiki

Stream Forge – Stories

Overview

The Stream Forge stories are short narrative scenarios used to motivate requirements and design decisions throughout Agile Semantic Knowledge Graph Design. Each story highlights a different class of problems commonly encountered when building semantic knowledge graphs in real-world systems.

These stories are not user stories or functional specifications. They are design narratives that explain why particular modeling choices, constraints, and metadata patterns are introduced.

For an overview of the Stream Forge example and its role in the book, see: → Stream-Forge

How the Stories Are Used

Across the book and supporting materials, the stories are used to:

  • Ground abstract concepts in realistic scenarios
  • Drive the definition of competency questions
  • Justify conceptual modeling decisions in OWL
  • Motivate validation and governance constraints in SHACL
  • Illustrate how semantic knowledge graphs support AI-assisted querying and reasoning

Each story introduces a slightly higher level of complexity than the previous one, mirroring how semantic systems evolve over time.

Story Index

Story 1 — Publishing and Identity

This story focuses on the ingestion of content metadata from multiple source systems and the challenge of establishing stable identity across heterogeneous identifiers.

Primary concerns:

  • Distinguishing identifiers from entities
  • Resolving duplicate or conflicting records
  • Establishing authoritative sources
  • Introducing basic provenance

This story motivates the foundational domain concepts and identity strategies used throughout the example.

→ Stream-Forge-Publishing-and-Identity

Story 2 — Metric Provenance and Trust

This story explores how derived metrics (such as engagement or popularity) are produced by different pipelines and why understanding their provenance is essential for trust and governance.

Primary concerns:

  • Tracking data lineage across transformations
  • Differentiating asserted vs computed values
  • Modeling activities, agents, and entities
  • Supporting audit and compliance requirements

This story drives the explicit modeling of provenance and governance metadata.

→ Stream-Forge-Metric-Provenance-and-Trust

Story 3 — Semantic Grounding for AI Responses

This story examines how an AI assistant answers analytical questions using a semantic knowledge graph and how the system communicates uncertainty, inference, and fallback behavior.

Primary concerns:

  • Semantic classification vs statistical inference
  • Evidence-backed responses
  • Transparent fallback to non-grounded LLM output
  • Reducing hallucination risk through semantic grounding

This story motivates the integration of semantic knowledge graphs with retrieval-augmented generation (RAG).

→ Stream-Forge-Semantic-Grounding-for-AI-Responses

Relationship to Other Materials

Each story links to:

  • Relevant requirements artifacts (competency questions, glossary entries)
  • Portions of the domain model introduced to address the story
  • Example data, queries, and validation rules that make the discussion concrete

Stories are intentionally reused across multiple sections of the book to demonstrate how early modeling decisions support later capabilities.

Status

The stories are stable conceptual anchors.
While supporting artifacts may evolve, the narratives themselves change only when necessary to preserve alignment with the published text.