Guidelines for the Annotation and Visualization of Legal Argumentation Structures in Chinese Judicial Decisions

This guideline proposes a systematic and operational annotation framework for representing the structure of legal argumentation in judicial decisions. Grounded in theories of legal reasoning and argumentation, the framework aims to reveal the logical organization of judicial reasoning and to provide a reliable data foundation for computational analysis. At the proposition level, the guideline distinguishes four types of propositions: general normative propositions, specific normative propositions, general factual propositions, and specific factual propositions. At the relational level, five types of relations are defined to capture argumentative structures: support, attack, joint, match, and identity. These relations represent positive and negative argumentative connections, conjunctive reasoning structures, the correspondence between legal norms and case facts, and semantic equivalence between propositions. The guideline further specifies formal representation rules and visualization conventions for both basic and nested structures, enabling consistent graphical representation of complex argumentation patterns. In addition, it establishes a standardized annotation workflow and consistency control mechanisms to ensure reproducibility and reliability of the annotated data. By providing a clear conceptual model, formal representation rules, and practical annotation procedures, this guideline offers methodological support for large-scale analysis of judicial reasoning and for future research in legal argument mining, computational modeling of legal reasoning, and AI-assisted legal analysis.

💡 Research Summary

The paper presents a comprehensive operational framework for annotating and visualizing the structure of legal argumentation in Chinese judicial decisions, specifically targeting the reasoning sections of judgments. Grounded in legal reasoning and argumentation theory, the guideline defines a clear conceptual model, formal representation rules, visualization conventions, an annotation workflow, and consistency control mechanisms to enable reproducible, large‑scale data creation for computational legal studies.

At the proposition level, four distinct types are introduced: General Normative judgments (GM), Specific Normative judgments (SM), General Factual judgments (GF), and Specific Factual judgments (SF). GM covers abstract legal norms such as statutes, principles, and precedents, and is further sub‑classified. SM refers to concrete normative provisions directly applicable to the case. GF denotes broad factual statements, while SF captures case‑specific factual details. The guideline stresses that proposition type assignment must rely on explicit textual cues, prioritize matching of normative‑factual pairs, and avoid duplication.

At the relational level, five relation types are defined: Support (S), Attack (A), Joint (J), Match (M), and Identity (I). Support and Attack encode positive and negative argumentative directions, respectively. Joint captures conjunctive reasoning where multiple premises are jointly required for a conclusion. Match describes the correspondence between a normative proposition and a factual proposition (e.g., a law applied to a fact). Identity denotes semantic equivalence between two propositions. For each relation, formal expression rules are provided (e.g., S: P → Q, A: P ↛ Q, J: P ∧ R → Q, M: Norm ↔ Fact, I: P ≡ Q), and the guideline distinguishes between non‑nested and multi‑level nested structures, prescribing hierarchical representation to preserve clarity.

Visualization standards are meticulously specified. Proposition nodes are rendered as rectangles; relation nodes as circles. Circle styles differentiate relation types: solid circles for Support, hollow circles for Attack, a “+” inside the circle for Joint, and a distinct style for Match. Identity is indicated by a slash “/” inside the rectangle. For nested structures, diagrams must be built layer‑by‑layer from the innermost relation outward, preserving independence of each relational node and ensuring readability.

The annotation workflow consists of five stages: (1) defining annotation scope and segmenting the document, (2) labeling proposition types, (3) assigning relation types, (4) constructing the argument diagram, and (5) performing quality assurance. Core principles guiding the workflow include prioritized matching (norm‑fact alignment first), non‑duplication, and reliance on explicit textual expressions. Consistency control mechanisms comprise double‑blind annotation, automated consistency‑checking scripts (verifying type‑relation coherence), and versioned guideline management to track updates and ensure reproducibility.

The guideline outlines four major impact areas. In legal education, visualized argument structures help students grasp the logical flow of judgments, linking facts, norms, and conclusions. For legal scholarship, the standardized dataset enables empirical comparison of reasoning across cases, judges, and courts, moving research from purely doctrinal analysis to data‑driven investigation. Regarding judicial transparency, the structured representation makes the reasoning behind decisions accessible to litigants and the public, facilitating quality assessment and supervision. Finally, in the development of judicial AI, the annotated structures provide a foundation for explainable models that can articulate reasoning paths, satisfy normative compliance, and be evaluated against explicit criteria of rationality and accountability.

Overall, the guideline establishes a logically rigorous, formally unified, and operationally practical standard for annotating and visualizing legal argumentation in Chinese judgments. By bridging doctrinal theory with computational methods, it supplies the essential data infrastructure for large‑scale argument mining, computational modeling of legal reasoning, and the creation of AI‑assisted tools that respect the explanatory demands of the rule of law.