Argument Annotation Schema

The token-level or span-level task of identifying and categorizing the functional parts of an argument within a sentence. This forms the foundational layer of any annotation schema.

• Premise: A statement providing a reason or evidence for a claim.
• Conclusion: The central proposition the author seeks to establish.
• Example: In 'The defendant was negligent because he ran a red light,' the schema labels 'he ran a red light' as a Premise and 'The defendant was negligent' as a Conclusion.

The binary or multi-class task of determining the rhetorical relationship between two annotated argument components. This transforms a flat list of components into a structured argument graph.

• Support: Component A strengthens or provides evidence for Component B.
• Attack: Component A weakens, contradicts, or rebuts Component B.
• Application: Essential for training models to perform Reasoning Chain Reconstruction and distinguish a coherent argument from a collection of unrelated statements.

A high-granularity annotation schema that decomposes legal arguments into six functional components defined by philosopher Stephen Toulmin. This provides a richer structure than simple premise-conclusion pairs.

• Claim: The assertion being argued for.
• Data: The facts or evidence grounding the claim.
• Warrant: The legal rule or principle connecting the data to the claim.
• Backing: The authority certifying the warrant (e.g., a statute).
• Qualifier: A phrase indicating the degree of certainty (e.g., 'likely').
• Rebuttal: A statement of an exception or counter-condition.

A sequence labeling task that classifies entire sentences in a legal judgment by their discourse function, distinct from the micro-level argument structure. This provides crucial document-level context.

• Common Roles: 'Facts of the Case,' 'Arguments by Petitioner,' 'Issue Framing,' 'Ratio Decidendi,' and 'Final Decision.'
• Synergy: When combined with Argumentative Zoning, this dual-layer annotation allows a model to learn both what a sentence is doing rhetorically and where it fits in the overall argumentative flow of a judicial opinion.

The classification of text spans expressing normative concepts of obligation, permission, and prohibition. This is critical for distinguishing factual arguments from legal rule statements.

• Obligation: 'must,' 'shall,' 'is required to.'
• Permission: 'may,' 'is entitled to,' 'has the right to.'
• Prohibition: 'must not,' 'shall not,' 'is barred from.'
• Purpose: This tagging enables the precise extraction of actionable legal rules for Deontic Logic Modeling and compliance checking systems.

The annotation process of identifying and connecting related argument components across multiple legal filings, such as linking a claim in a complaint to its counter-argument in a motion to dismiss.

• Entity Coreference: Linking mentions of the same party or object across documents.
• Argument Coreference: Linking the same legal claim as it is addressed, rebutted, or refined in subsequent briefs.
• Goal: This schema extension is fundamental for building Multi-Document Legal Reasoning systems that can synthesize a complete case narrative from a docket of filings.

The token-level or span-level task of identifying and categorizing the functional parts of an argument, such as a premise or a conclusion, within a sentence. An annotation schema provides the exact label set for this task.

• Premise: A reason or piece of evidence supporting a claim.
• Conclusion: The central assertion the arguer seeks to prove.
• Major Claim: The ultimate thesis of a document.

The binary or multi-class task of determining whether one legal argument component strengthens, weakens, or is neutral toward another. Annotation schemas define the directed edges in an argument graph.

• Support: Premise A provides a reason to accept Claim B.
• Attack: Premise A provides a reason to reject Claim B.
• Undercut: An attack that targets the inferential link between a premise and a conclusion.

The decomposition of legal arguments into the six functional components defined by philosopher Stephen Toulmin. A Toulmin-based annotation schema is one of the most influential frameworks for capturing the deep structure of reasoning.

• Claim: The conclusion being argued for.
• Data: The evidence or facts grounding the claim.
• Warrant: The rule or principle linking data to the claim.
• Backing: The authority supporting the warrant.
• Qualifier: The degree of certainty (e.g., 'probably').
• Rebuttal: The conditions under which the claim would not hold.

A technique for segmenting a legal document into distinct rhetorical blocks based on the author's purpose. This is a coarser annotation task that often precedes fine-grained argument mining.

• Aim: Distinguishes argumentation from background exposition, procedural history, or dicta.
• Use Case: Provides structural context for downstream component classification.
• Granularity: Operates at the sentence or paragraph level.

A foundational mathematical framework that models arguments as abstract nodes in a directed graph, focusing solely on attack relations to determine acceptable sets of claims. Annotation schemas provide the concrete instances for these abstract frameworks.

• Abstract: Ignores the internal structure of arguments.
• Semantics: Defines rules (e.g., grounded, preferred) for computing which arguments are collectively acceptable.
• Output: Produces sets of winning arguments from a dispute.

The algorithmic assembly of individual argument components into a coherent, step-by-step inferential path from legal premises to a final conclusion. This task relies on a schema that captures transitive relations.

• Goal: Build a connected path: Premise A → Intermediate Claim B → Final Conclusion C.
• Challenge: Requires resolving implicit or missing premises (enthymemes).
• Application: Essential for generating explainable legal AI outputs.