luvv to helpDiscover the Best Free Online Tools
Topic 1 of 8

Creating Domain Glossaries And Rules

Learn Creating Domain Glossaries And Rules for free with explanations, exercises, and a quick test (for Prompt Engineer).

Published: January 8, 2026 | Updated: January 8, 2026

Table of Contents

Why this matters

LLMs pick words statistically, not by your domain standards. A domain glossary and a ruleset turn fuzzy language into consistent, safe outputs. As a Prompt Engineer, you will often:

  • Define preferred terms and forbid confusing synonyms (e.g., patient vs. client).
  • Standardize units, dates, and capitalization (e.g., mmHg, ISO dates).
  • Set tone and structure (e.g., short bullet summaries first, then details).
  • Redact sensitive information (e.g., remove emails, phone numbers, IDs).
  • Add disclaimers or required phrases.
  • Create replacement rules for common errors (e.g., glucose units mg/dL, not mg/mL).

Who this is for

  • Prompt Engineers and Applied ML/AI professionals who need reliable, domain-accurate outputs.
  • Technical writers and product managers building LLM features for specific industries.

Prerequisites

  • Basic prompt design skills (system vs. user prompts, few-shot examples).
  • Familiarity with the domain you are targeting, or access to a subject-matter expert (SME).
  • Comfort reading small corpora (docs, tickets, emails) to extract terms.

Concept explained simply

A domain glossary is a living list of terms with preferred wording and examples. Rules are enforceable instructions that shape style, formatting, safety, and structure. Together, they act like a language contract for the model.

Mental model: Three gates for every output

  1. Vocabulary Gate: Does the output use the right terms and definitions?
  2. Format Gate: Are units, dates, numbers, and capitalization standardized?
  3. Policy Gate: Is the content safe, on-brand, and compliant (redaction, disclaimers)?

How to build a domain glossary and rules

  1. Collect raw language: samples of real tasks, tickets, docs, and user intents.
  2. List candidates: domain nouns, verbs, abbreviations, units, names, and forbidden terms.
  3. Decide preferred forms: pick one label per concept; add synonyms to map to it.
  4. Write crisp definitions: one-sentence, user-facing, with do/don’t examples.
  5. Add formatting policies: dates, numbers, currency, units, capitalization.
  6. Add guardrails: redaction, banned phrases, mandatory disclaimers.
  7. Test with real prompts: run 10–20 tasks; note failures; refine entries and rules.
  8. Version and track changes: mark date, owner, and reason for changes.
Glossary entry template (copy-friendly)
  • Term: Preferred label
  • Definition: One clear sentence
  • Synonyms map to: [list]
  • Forbidden terms: [list + reason]
  • Examples (good/bad): 1–2 short lines
  • Formatting: capitalization, plural, unit or symbol
  • Notes: domain-specific nuance
Ruleset template (copy-friendly)
  • Style: tone (e.g., concise, confident), voice (active), and structure (summary first).
  • Formatting: dates (YYYY-MM-DD), numbers (1,234.56), currency (USD), units (SI).
  • Vocabulary enforcement: use preferred terms; auto-replace synonyms.
  • Safety: redact PII/PHI patterns; avoid giving medical/financial advice if prohibited.
  • Disclaimers: append required statements when specific topics occur.
  • Validation & fallback: if unknown term, ask a clarifying question instead of guessing.

Worked examples

1) Healthcare telemedicine notes

  • Glossary snippet:
    • Term: Blood pressure (BP). Definition: Arterial pressure measured in mmHg.
    • Synonyms map: bp, b.p., bloodpressure → Blood pressure (BP)
    • Formatting: Use mmHg; show as systolic/diastolic (e.g., 120/80 mmHg).
    • Forbidden: “tension” (ambiguous).
  • Rules:
    • Units: Use SI; BP in mmHg only; glucose mg/dL.
    • Safety: Remove names, phone numbers, account IDs from outputs.
    • Structure: Summary (1–2 bullets), then Observations, then Recommendations.
    • Disclaimer: Add “This summary is informational and not a diagnosis.”

2) E-commerce fashion product pages

  • Glossary snippet:
    • Term: SKU. Definition: Internal stock-keeping unit identifier.
    • Synonyms map: item code, stock id → SKU
    • Term: Care instructions. Definition: Steps for washing/drying/ironing.
    • Forbidden: “100% guarantee” (legal risk).
  • Rules:
    • Tone: Positive and factual; no superlatives without evidence.
    • Sizes: Use EU/US conversions consistently; include a size guide note.
    • Formatting: Title Case for product names; bullet lists for features.
    • Safety: Remove emails/phone numbers from seller notes.

3) Financial risk reports

  • Glossary snippet:
    • Term: Value at Risk (VaR). Definition: Maximum expected loss over a period at a confidence level.
    • Synonyms map: value-at-risk → Value at Risk (VaR)
    • Formatting: Show confidence as %, horizon in days.
    • Forbidden: “guaranteed returns”.
  • Rules:
    • Numbers: Use two decimals; thousands separators.
    • Dates: YYYY-MM-DD.
    • Structure: Executive summary → Key metrics table (described in text) → Risks → Notes.
    • Compliance: Add “Not investment advice.” if recommendations appear.

Common mistakes and self-check

  • Mistake: Vague definitions. Fix: One-sentence, test with a non-expert.
  • Mistake: Too many synonyms left ungrouped. Fix: Map all to a single preferred term.
  • Mistake: Inconsistent units/dates. Fix: Declare one canonical format.
  • Mistake: Missing safety rules. Fix: Add redaction + disclaimers for risky topics.
  • Mistake: Never testing on real tasks. Fix: Validate with 10–20 real prompts and iterate.

Self-check before rollout:

  • Can a new team member use the glossary without asking questions?
  • Do 95%+ outputs follow units, dates, and tone automatically?
  • Are sensitive fields reliably redacted in tests?

Step-by-step mini workflow

  1. Extract 30–100 candidate terms from real docs.
  2. Cluster into 10–25 core concepts with preferred labels.
  3. Write definitions and examples; specify forbidden terms.
  4. Add rules for tone, structure, formats, and safety.
  5. Test on representative tasks; note deviations; update rules.

Exercises

Do these hands-on tasks. The Quick Test at the end checks your understanding. Everyone can take the test; logged-in users will see saved progress.

Exercise 1: Draft a mini glossary and rules for an EdTech Q&A bot

Task: Create a 10–12 term glossary for a university EdTech Q&A bot (topics: enrollment, tuition payments, deadlines). Add 6–8 rules for tone, structure, dates, and safety.

  • Include preferred terms, synonyms mapping, and forbidden terms with reasons.
  • Define date format and sensitive info redaction.

Expected output: A compact glossary list and a numbered ruleset.

Hints
  • Prefer “tuition” over “school fees”; map synonyms.
  • Use YYYY-MM-DD; redact student ID, email, phone.
Show solution

Sample solution (abridged):

  • Tuition. Definition: Cost charged per term for enrollment. Synonyms: school fees → Tuition. Forbidden: “price” (too generic).
  • Enrollment. Definition: Officially registering for courses. Synonyms: sign-up, registration → Enrollment.
  • Financial aid. Definition: Grants, scholarships, or loans that reduce tuition.
  • Registrar. Definition: Office managing academic records and registration.
  • Student ID. Definition: Unique identifier for a student. Formatting: 8 digits. Safety: redact in outputs.
  • Deadline. Definition: Last acceptable date for an action. Formatting: YYYY-MM-DD.
  • Waitlist. Definition: Queue for course seats when full.
  • Prerequisite. Definition: Course required before another course.
  • Semester. Definition: Academic period (Fall, Spring, Summer).
  • Transcript. Definition: Official academic record.
  1. Tone: Helpful, neutral, concise; use active voice.
  2. Structure: Summary (1–2 lines) → Steps → Notes.
  3. Dates: Use YYYY-MM-DD (e.g., 2026-09-01).
  4. Numbers: Use thousands separators and USD for currency.
  5. Safety: Redact emails, phone numbers, and Student IDs.
  6. Vocabulary: Use preferred terms; replace “school fees” → “tuition”.
  7. Fallback: If a policy differs by campus, ask a clarifying question.
Exercise 2: Create normalization rules for units and dates

Task: Define 6–10 normalization rules for a hardware support bot. Topics: measurements (cm/in), weights (kg/lb), timestamps, product codes.

Expected output: A numbered list of rules with examples.

Hints
  • Pick one canonical unit, and always show conversions in parentheses.
  • Use 24-hour time and ISO dates.
Show solution
  1. Lengths: Canonical cm; if input in inches, show both (e.g., 25.4 cm (10 in)).
  2. Weight: Canonical kg; show lb in parentheses with one decimal.
  3. Temperature: Canonical °C; if °F provided, show both with one decimal.
  4. Dates: YYYY-MM-DD only; convert from other formats.
  5. Time: 24-hour, local timezone noted (e.g., 16:30 local).
  6. Product code: Uppercase, hyphenated (e.g., ZX-120B); strip spaces.
  7. Decimals: Use period as decimal separator; thousands separators for numbers ≥ 1000.
  8. Ranges: Use en dash with spaces (e.g., 5 – 10 cm).

Build-and-review checklist

  • 10–25 core terms with definitions and examples.
  • Synonyms and forbidden terms mapped for each concept.
  • One canonical format for dates, numbers, currency, and units.
  • Safety rules: redaction + required disclaimers.
  • Structure and tone rules with 1–2 examples.
  • Validated on 10–20 real tasks; issues logged and fixed.

Practical projects

  • Create a glossary and ruleset for a customer support chatbot in cybersecurity (incident, alert, severity, playbook). Validate on 15 tickets.
  • Standardize product specs for an IoT catalog; normalize units and add safety (remove device serials).
  • Build a ruleset for internal meeting summaries: agenda-first, decisions, action items with owners and due dates.

Learning path

  1. Start: Collect corpus → extract terms → draft glossary entries.
  2. Add rules: tone, structure, formatting, and safety.
  3. Prototype: Run samples through your prompts using the glossary + rules.
  4. Evaluate: Compare outputs against checklist; iterate.
  5. Scale: Version the document; share with team; schedule updates.

Mini challenge

Pick any public domain (e.g., travel itineraries). Write 5 glossary entries with synonyms/forbidden terms and 5 rules. Test on two sample prompts. Note one improvement you’d make after reviewing outputs.

Next steps

  • Integrate your glossary and rules into your system or prompts as system instructions.
  • Schedule a monthly review with an SME to update entries and rules.
  • Proceed to the next subskill after passing the Quick Test.

Quick Test

The Quick Test is available to everyone. Logged-in users will see saved progress.

Practice Exercises

2 exercises to complete

Instructions

Create a 10–12 term glossary for a university EdTech Q&A bot (enrollment, tuition, deadlines). Add 6–8 rules for tone, structure, dates, and safety. Include synonyms mapping and forbidden terms with reasons.
Expected Output
A compact glossary (10–12 terms with synonyms/forbidden lists) and a numbered ruleset (6–8 items) covering tone, structure, formats, and safety.

Creating Domain Glossaries And Rules — Quick Test

Test your knowledge with 6 questions. Pass with 70% or higher.

6 questions70% to pass
Style Guides And Brand Voice

Have questions about Creating Domain Glossaries And Rules?

AI Assistant

Ask questions about this tool