Ambiguity Rubric#

A 5-dimension rubric for quantitatively evaluating ambiguity in spec documents. Final score = weighted average (0.0 = perfectly clear, 1.0 = completely ambiguous)

Pass Criteria#

Dimension 1: Lexical Precision — Weight 0.25#

Whether ambiguous words/expressions have been replaced with specific values or definitions.

Prohibited Word List (Ambiguous Expressions)#

Locale-tagged. The lexical scan MUST match the spec's own language. A Korean spec is scanned with the Korean column; do NOT rely on the English column alone (that misses every Korean vague term and silently scores 0.0 → see the locale rule below).

Scoring Criteria#

Locale coverage rule (unverified ≠ clear): If the spec's primary language is NOT covered by the prohibited-word table above, this dimension is UNVERIFIED, not 0.0. A keyword miss in an unsupported language must NOT be scored as "0 prohibited words" — that silently pulls the weighted total toward CLEAR/Pass (the LANG-01 fail-open). Mark the dimension UNVERIFIED, surface a warning, and fall back to LLM semantic judgment of vagueness rather than keyword counting.

Dimension 2: Reference Clarity — Weight 0.20#

Whether pronouns, demonstratives, and abbreviations clearly point to specific entities.

Inspection Items#

• Pronouns: "this", "that", "said" → Replace with specific entity names
• Abbreviations: Define full name + abbreviation on first occurrence
• Implicit subjects: "is processed", "is sent" → Specify the agent ("the server processes")
• External references: "as per existing approach" → Specify reference document + section

Scoring Criteria#

Dimension 3: Completeness — Weight 0.20#

Whether all features have acceptance criteria (AC) and edge cases have been identified.

Inspection Items#

• Acceptance criteria: Given-When-Then AC exists for all features/requirements
• Edge cases: Boundary values, empty inputs, concurrency, timeouts identified
• Error scenarios: Failure behavior defined (retry, fallback, alerts)
• Non-functional requirements: Performance, security, accessibility criteria specified

Scoring Criteria#

Dimension 4: Consistency — Weight 0.20#

Whether there are no contradictions between sections and terminology is unified.

Inspection Items#

• Terminology unity: Same term used for same concept (glossary exists)
• Numeric consistency: Same values consistent throughout the document
• Logical consistency: Premises in section A and conclusions in section B do not contradict
• Priority consistency: Must/Should/May consistent across sections

Scoring Criteria#

Dimension 5: Testability — Weight 0.15#

Whether all requirements can be verified through automation or manual testing.

Inspection Items#

• Measurable: Pass/Fail criteria exist for all requirements
• Reproducible: Same conditions yield same results during test execution
• Independent: Each requirement can be tested individually
• Time-bound: Time limit for test completion specified (when applicable)

Scoring Criteria#

Final Score Calculation#

ambiguity_score = (lexical × 0.25) + (reference × 0.20) + (completeness × 0.20)
                + (consistency × 0.20) + (testability × 0.15)

Report Format#

## Ambiguity Evaluation Results

| Dimension | Score | Weight | Weighted Score | Key Findings |
|-----------|-------|--------|---------------|-------------|
| Lexical Precision | {0.0-1.0} | 0.25 | {calculated} | {findings} |
| Reference Clarity | {0.0-1.0} | 0.20 | {calculated} | {findings} |
| Completeness | {0.0-1.0} | 0.20 | {calculated} | {findings} |
| Consistency | {0.0-1.0} | 0.20 | {calculated} | {findings} |
| Testability | {0.0-1.0} | 0.15 | {calculated} | {findings} |
| **Total** | | | **{sum}** | **{verdict}** |

### Improvement Recommendations
1. {Specific improvement item}