Deontic Annotation Schema

The primary label assigned to a text span indicating its normative force. This is the foundational layer of the schema.

• Obligation: Text imposing a duty (e.g., shall, must, is required to).
• Permission: Text granting a right or privilege (e.g., may, is entitled to, has the right to).
• Prohibition: Text forbidding an action (e.g., shall not, must not, is prohibited from).
• No Normative Force: Text that is purely definitional, descriptive, or informational.

Identifies the legal actors bound by or benefiting from the deontic modality. This disambiguates the direction of the normative relationship.

• Bearer: The entity upon whom the obligation or prohibition falls.
• Counterparty: The entity to whom the duty is owed or the permission is granted.
• Example: In 'The Licensee shall pay Licensor', the Bearer is 'Licensee' and the Counterparty is 'Licensor'.

Captures the antecedent conditions that activate, suspend, or terminate a deontic modality. This moves the annotation from static rules to dynamic, state-dependent norms.

• Activation Condition: The event or state that makes the obligation effective (e.g., upon delivery).
• Termination Condition: The event that discharges the duty (e.g., until full payment).
• Exception: A specific scenario where the general norm does not apply (e.g., except in cases of force majeure).

The specific conduct or subject matter governed by the deontic modality. This is the verb phrase or nominal clause that constitutes the regulated behavior.

• Action Span: The text describing the required, permitted, or prohibited act (e.g., deliver the source code, use the trademark).
• Temporal Constraints: Deadlines, notice periods, or frequency adverbs attached to the action (e.g., within 30 days, annually).
• Manner Specifications: Qualitative requirements on how the action must be performed (e.g., in writing, using commercially reasonable efforts).

Metadata labels that identify when a deontic annotation is in tension with another norm and how the schema resolves precedence.

• Conflict ID: A shared identifier linking two or more annotations that prescribe incompatible actions.
• Resolution Principle: A tag indicating the meta-rule used to resolve the conflict (e.g., lex specialis, lex posterior, lex superior).
• Defeasibility Marker: A flag indicating that the norm is subject to being overridden by a more specific or higher-priority rule.

Links each annotation back to its authoritative source within the document structure, ensuring full traceability for legal verification.

• Source Span: The exact character offsets of the text justifying the annotation.
• Section Reference: The structural element (e.g., Section 4.2(a)) from which the norm originates.
• Confidence Score: A numerical value indicating annotator or model certainty, critical for managing ambiguity in gold-standard corpora.

A fundamental analytical framework for disambiguating legal relations before annotation. It decomposes vague 'rights' into precise jural correlatives:

• Right/Duty: A claim one party has against another.
• Privilege/No-Right: A freedom from a duty.
• Power/Liability: The ability to change a legal relation.
• Immunity/Disability: Freedom from having a relation changed. Applying this analysis ensures the schema captures the exact legal posture of each actor.

A critical annotation target representing fallback obligations triggered by a violation. A robust schema must capture these conditional structures to model real-world legal complexity. For example, annotating a primary obligation ('pay tax by April 15') and its CTD counterpart ('pay tax with a penalty after April 15') is essential to avoid the logical contradictions of Chisholm's Paradox.

The schema's mechanism for tagging situations where two or more applicable norms prescribe incompatible actions. Annotators must identify and link conflicting obligations and apply resolution tags based on legal principles:

• Lex Superior: Higher authority prevails.
• Lex Specialis: The more specific rule prevails.
• Lex Posterior: The later-enacted rule prevails. This creates training data for conflict resolution models.

An OASIS standard for legal rule interchange that directly informs schema design. Aligning your annotation schema with LegalRuleML's semantic categories ensures interoperability with other normative reasoning systems. It provides a standardized XML-based structure for encoding the deontic tags, actors, and conditions identified during the annotation process.

The primary NLP benchmark task for evaluating models trained on this schema. DTE determines if a textual premise normatively entails a deontic conclusion. A high-quality annotation schema is the prerequisite for creating a gold-standard DTE dataset, where each premise-hypothesis pair is meticulously labeled with obligation, permission, or prohibition relations.