ZALARI Convention

Zero-loss Abridged Legal Annotation for Reasoning Interfaces v1.0

The ZALARI Convention establishes a universal citation format for encoding legal authorities in AI reasoning contexts. Designed for use in large language model prompts, retrieval-augmented generation pipelines, and structured legal AI systems, the Convention prioritises token efficiency, parse determinism, and semantic fidelity above conventional typographic form.

This document specifies Version 1.0 of the ZALARI format and serves as the canonical reference for all ZALARI-compliant implementations. The Convention is maintained by DeepCounsel and governs the encoding of case law and statutory authorities across all supported jurisdictions.

Core Specification

The canonical ZALARI citation format is:

[TITLE], [CITATION_ID], ([COUNTRY]), [PINPOINT]

This format is governed by the following design constraints:

• No full stops within abbreviations
• No italic markers or typographic decoration
• No redundant reporter prefixes where neutral citations are available
• Consistent delimiters using comma-space throughout
• Country code in parentheses immediately following the citation identifier

Component Definitions

TITLE — The case name as it appears in the official reporter or neutral citation, preserving the original party names without abbreviation. Party names must not be shortened, initialled, or abbreviated in any form. Where a case is known by a short name in legal practice, the full formal title governs in ZALARI encoding.

CITATION_ID — The primary neutral citation where available, or the most-cited reporter citation where a neutral citation has not been assigned. Selection rule: prefer neutral citation over reporter citation; where multiple reporters exist, prefer the official or government reporter. Example: 2019 ZWSC 4

COUNTRY — The ISO 3166-1 alpha-3 country code in parentheses. Examples: (ZWE), (ZAF), (CAN), (GBR), (USA)

PINPOINT — Paragraph, page, or section reference using the abbreviated ZALARI pinpoint syntax. Format: p[N] for paragraph number, pg[N] for page number, s[N] for section number.

Punctuation Austerity

The ZALARI Convention eliminates conventional legal punctuation that does not contribute semantic content. The following transformations apply universally:

• R. v. → R v
• S.C.C. → SCC
• U.S. → US
• at para. → p
• [2019] (year bracket) retained where it forms part of the official citation
• All italic markers removed
• All period-terminated abbreviations converted to period-free form

Implementation Rationale

Token Efficiency — Legal citation strings are high-frequency content in legal AI systems. Reducing token count per citation by 15–30% through punctuation austerity materially reduces inference cost and context window pressure in production legal AI deployments.

Parse Determinism — The fixed delimiter structure (TITLE, CITATION_ID, (COUNTRY), PINPOINT) enables deterministic regex and grammar-based parsing without ambiguity. This supports reliable extraction, indexing, and verification of citations in automated legal processing pipelines.

Vector Similarity — Consistent formatting produces stable embeddings for citation strings. Normalised forms cluster correctly in semantic vector spaces, enabling reliable retrieval of related authorities across jurisdictions.

Jurisdiction Specifications

Canada — Country code: (CAN). Neutral citations preferred where issued by the court. Abbreviations: SCC (Supreme Court of Canada), ONCA (Ontario Court of Appeal), BCCA (British Columbia Court of Appeal). Party name R used without periods for Crown prosecution.

United States — Country code: (USA). Federal citations follow US Reports or Federal Reporter format without periods. State citations use standard state abbreviations without periods. Example: Miranda v Arizona, 384 US 436, (USA), p478

United Kingdom — Country code: (GBR). Neutral citations from UKSC, EWCA, EWHC preferred. Law Reports citations used where neutral citation unavailable. Example: Donoghue v Stevenson, [1932] AC 562, (GBR), p580

Zimbabwe — Country code: (ZWE). Full specifications in ZALARI 1.0.

Zimbabwean Case Law Specifications

Zimbabwean case law presents unique encoding requirements arising from the formal nature of party designations in the Zimbabwean legal tradition. The following rules govern ZALARI encoding of Zimbabwean authorities and are binding on all ZALARI-compliant implementations.

Full Names Requirement — All party names in Zimbabwean citations must be encoded in full. No abbreviation, initialisation, or truncation of party names is permitted. This requirement reflects the formal naming conventions of Zimbabwean courts and ensures that citations can be reliably matched against official court records.

Multiple Parties Notation — Where a case involves multiple parties on one side, ZALARI uses the following conventions:

• & Anor — one additional party
• & 2 Ors, & 3 Ors, etc. — multiple additional parties (numeral + Ors)

State Prosecution — In criminal proceedings brought by the state, the prosecuting authority appears as The State (not R or Regina or Rex). The State appears as the first-named party.

The State v Chikomo, 2018 ZWSC 11, (ZWE), p4
The State v Mhiripiri & Anor, 2020 ZWHHC 145, (ZWE), p7
The State v Ndlovu & 2 Ors, 2019 ZWSC 22, (ZWE), p3
Telecel Zimbabwe (Pvt) Ltd v Postal and Telecommunications Regulatory Authority of Zimbabwe & Anor, 2016 ZWSC 5, (ZWE), p12

Migration Protocols

Systems migrating from conventional citation formats to ZALARI should apply transformations in the following order: (1) remove italics and typographic markers; (2) strip periods from abbreviations; (3) apply jurisdiction-specific party name rules; (4) insert country code; (5) convert pinpoint references to ZALARI pinpoint syntax; (6) validate against ZALARI schema.

Batch migration tools and validation utilities are available through the Zalari developer API for organisations migrating existing citation databases to ZALARI format.

Conclusion

ZALARI solves three persistent problems in legal AI engineering:

• Token waste from conventional punctuation in high-volume citation contexts
• Parse ambiguity arising from inconsistent formatting across jurisdictions
• Embedding instability from orthographic variation in citation strings

By establishing a single, deterministic, jurisdiction-aware encoding standard, ZALARI enables legal AI systems to handle citations with the same precision and efficiency as structured data, while preserving the semantic integrity required for reliable legal reasoning.

For queries about the ZALARI Convention or to report implementation issues, contact legal@deep-counsel.org.