LIRIL Classification System

What is LIRIL?

LIRIL is the ethics gate and classification engine of the TENET5 system. Named for the palindromic center of the SATOR square (SATOR → AREPO → TENET → OPERA → ROTAS), LIRIL processes every input through a deterministic classification pipeline before any other agent sees it.

In the context of The 504 Database, LIRIL classifies each government accountability record into one of six categories based on keyword analysis, source verification, and severity assessment. The classification results appear as banners on investigation pages and as data attributes on individual records.

Input Record
    ↓
┌─────────────────────────────┐
│  1. Keyword extraction      │
│  2. Source verification     │
│  3. Severity scoring        │
│  4. Category assignment     │
│  5. Confidence calculation  │
│  6. ARTSTEM routing         │
└─────────────────────────────┘
    ↓
Classification Result
    ↓
Internal message routing
    ↓
Build-time bake → HTML data attributes
        

The SATOR Square

LIRIL's routing is based on the SATOR square — a 5×5 Latin palindrome that reads the same in all directions. In the TENET5 architecture, each row of the SATOR square maps to a routing function:

LIRIL operates at Row 3 — TENET — the center axis. It is the pivot point between ingestion (rows 1–2) and output (rows 4–5). Every classification decision passes through the TENET row. The LIRIL bar on each page shows SATOR·TENET·ROTAS — indicating data was sown (SATOR), worked (OPERA), and classified at the center (row3 = TENET).

Classification Categories

LIRIL classifies each record into one of six categories. Classifications are deterministic — the same input always produces the same output given the same SEED value (118400).

Record counts as of last build. Total: 1,105 records across all categories.

Classification Methodology

Each record passes through a 5-step classification pipeline:

Build-Time Baking

LIRIL classifications are baked at build time, not computed on page load. This is a deliberate architectural choice:

• Reproducibility: Given the same SEED (118400) and the same input data, LIRIL produces identical classifications every time
• Auditability: Every classification result is committed to the Git repository as static HTML. The full classification history is in version control
• No runtime dependency: The site works without any backend, API, or live connection to the TENET5 mesh. tenet5.liril.classify is used during the build process, not at page load
• Tamper evidence: Because classifications are in Git, any modification is visible in the commit history. You cannot change a record's classification without creating a Git commit

Records (HTML) → LIRIL classify (tenet5.liril.classify)
    ↓
Internal message routing
    ↓
HTML data attributes written:
  data-type="conviction|charge|ethics|scandal|expense|conflict"
  data-level="federal|provincial|municipal"
  data-party="lib|con|ndp|ind"
    ↓
Git commit → Git push → GitHub Pages deploy
    ↓
Static HTML served — no runtime classification
        

Internal Communication

LIRIL communicates over an internal NATS message bus. The following subjects are used during the build process:

Transparency Commitment

The 504 Database is built on the principle that government accountability records should be publicly accessible, machine-readable, and independently verifiable. LIRIL's classification system is documented here because:

• No classification is hidden. Every record's category is visible in the HTML source as data-type attributes.
• No algorithm is undisclosed. The keyword-decisive method is documented on this page. The SEED is public.
• No record is unsourced. Every record in the database links to primary or secondary sources — AG reports, court records, Hansard, committee testimony, or investigative journalism with cited primary documents.
• Version history is public. The TENET-5.github.io repository is open. Every classification change is a Git commit.
• Section 504 is the foundation. Criminal Code s.504 gives every Canadian the right to lay an information before a justice. This database provides the documented evidence base for citizens to exercise that right.