error · Section 1: Words
Rule 1.2: Use approved words as the specified part of speech
Use an approved dictionary word only as the part of speech (noun, verb, adjective, etc.) that the dictionary specifies for it; do not convert it to a different part of speech.
What a checker looks for
A clear deviation from the rule.
- Flag a word approved as one part of speech (e.g. a noun or adjective) being used as a different part of speech (e.g. as a verb) — for example 'test' is an approved noun but not an approved verb, so 'test the system' is a violation.
- Flag adjectives used as verbs (e.g. 'dim the lights') when the dictionary only approves the adjective form.
What this rule does not cover
- A word approved as more than one part of speech (e.g. 'clean' as both verb and adjective) is compliant either way; its position in the sentence shows which is being used.
How to fix it
Reconstruct the sentence so the word is used only in its approved part of speech — often this means turning a verb usage into a noun phrase ('do the leak test of the system' instead of 'test the system') or finding an approved verb with the same meaning ('make sure the valve can operate' instead of 'make sure the valve is operable').
Constraints on the fix
- Preserve the original instruction's meaning exactly.
- If the straightforward replacement changes meaning, use a different word or sentence construction instead.
Examples
Not STE: Test the system for leaks.
STE: Do the leak test of the system.
Not STE: Dim the lights.
STE: Set the lights to the dim position.
Related rules
Where this fits in AI governance
A rule a checker can test is a rule a model can be held to. The same discipline, on the governance side of this site.
- Hallucination
An approved word used in an unapproved sense is the controlled-language version of a plausible wrong answer. - Misinformation and confabulation
The register entry for text that reads correctly and is not. - Output validation guardrails
A controlled language is a checkable output constraint. This is the control that treats model output as something to be checked rather than trusted.
Source
ASD-STE100 Issue 9, Section 1 - Words, page 1-1-2.
This page is commentary and does not reproduce the standard. About this rule pack.