Skilirium

Document data rules as SBVR fact types and rule statements

Activity 3 of 8 in Data modelling · Occupation: Data modeller

The outcome: documents data rules in SBVR and resolves rule anomalies.

Builds on: Define business terms.

Which AI model is best at document data rules as sbvr fact types and rule statements?

Best result for document data rules as sbvr fact types and rule statements: claude-sonnet-5-5 (95% test score).

ModelProviderTierTest scoreKnobits mastered (expected)Cost per test run
gemini-4-argonGoogleFlagship86% 97%$0.062
gemini-3.1-pro-previewGoogleFlagship74% 86%$0.20
gpt-6-astraOpenAIFlagship89% 94%$0.45
claude-fable-5-1AnthropicFlagship88% 98%$0.85
gemini-3.8-flashGoogleMid tier82% 93%$0.045
gpt-6.1-solOpenAIMid tier88% 94%$0.077
claude-sonnet-5-5AnthropicMid tier95% 100%$0.23
gemini-3.5-flash-liteGoogleLow cost35% 53%$0.008
gpt-6-lunaOpenAILow cost76% 87%$0.005
claude-haiku-5-5AnthropicLow cost88% 93%$0.018

Test score: mean rubric score on the test tasks for this activity. Knobits mastered: expected share of the activity's knobits the model masters. Cost: mean API cost of one test run at list prices. Method and caveats.

The 28 knobits of this activity

25 that AI does well, 0 it does partly, and 2 where people still lead.

IDKnobitTypeAI exposure (models mastering)
K-301Recognises contradicting data rulesSkillAI does this well 7/10
K-302Recognises a data rule no valid data can satisfySkillNot tested yet
K-303Recognises redundant data rulesSkillAI does this well 9/10
K-304Recognises a data rule implied by anotherSkillAI does this well 9/10
K-305Recognises overlapping value-set rules on one fact typeSkillAI does this well 7/10
K-306Recognises circular derivation rulesSkillAI does this well 7/10
K-307Recognises a fact type with clashing modalitiesSkillAI does this well 9/10
K-308Recognises one relationship stated as two fact typesSkillAI does this well 7/10
K-309Recognises a gap in a partition of casesSkillPeople still lead 1/10
K-310Recognises a constraint claimed by a rule and missing from the schemaSkillAI does this well 10/10
K-311Recognises a uniqueness rule with no unique key behind itSkillAI does this well 8/10
K-312Recognises a rule assigned to a missing data quality checkSkillAI does this well 9/10
K-313Recognises a declared constraint the target platform does not enforceSkillAI does this well 10/10
K-314Recognises a data-update rule with no enforcing or detecting mechanismSkillAI does this well 8/10
K-315Recognises an activity or party rule claimed as a data ruleSkillAI does this well 9/10
K-316Recognises an ambiguous pronoun in a rule statementSkillAI does this well 9/10
K-317Recognises mixed and-or without precedence in a ruleSkillAI does this well 9/10
K-318Recognises a cardinality rule stated from one end onlySkillAI does this well 9/10
K-319Recognises a definitional rule filed as operativeSkillAI does this well 9/10
K-320Recognises an operative rule filed as definitionalSkillAI does this well 9/10
K-351Explains SBVR fact types and rule statementsKnowledgeAI does this well 9/10
K-352Explains alethic and deontic modality of rulesKnowledgeAI does this well 9/10
K-353Explains logical relations between rulesKnowledgeAI does this well 9/10
K-354Explains derivation rules and their dependenciesKnowledgeAI does this well 7/10
K-355Explains Witt's classification of business rulesKnowledgeAI does this well 10/10
K-356Explains rule enforcement mechanisms and platform enforcementKnowledgeAI does this well 10/10
K-358Explains partitions of a concept into casesKnowledgePeople still lead 1/10
K-337Resolves contradicting rules by adding the missing conditionSkillAI does this well 7/10