Can controlled languages be based on languages other than English?

Yes. English is common in examples, but the concept is not limited to English.

Does controlled language remove every ambiguity?

No. Some designs deliberately preserve flexibility. Others fix particular interpretation rules, but even then domain vocabulary and intended meaning require care.

AI Glossary / Natural language and speech / Intermediate

Controlled natural language

Definition

Definition: A controlled natural language is a deliberately designed form of a natural language with specified restrictions on vocabulary, grammar or interpretation. Its purpose can be clearer human communication, easier translation or more precise computer processing.

What it means

What it means: Ordinary language allows many ways to express an idea and many potential ambiguities. A controlled language narrows those possibilities for a defined purpose. It may require approved terms, particular sentence patterns or explicit conventions for interpreting a statement.
There is no single universal controlled language. A version intended for technical documentation can differ greatly from one designed to map sentences into formal logic. Some prioritize readability; others prioritize precision. Restrictions alone do not guarantee that readers understand the text or that its factual claims are correct. The language's actual rules and intended use need to be identified.

How it works

How it works: Designers define permissible expressions and explain how authors should use them. A controlled vocabulary can fix preferred terms, while grammatical rules limit sentence constructions. Interpretation rules may resolve scope or reference in a specified way. Authoring tools can offer permitted continuations or check whether text follows the defined language.
For example, Attempto Controlled English publishes construction and interpretation rules intended to support a precise formal reading. That is one particular design, not a property of every controlled language. A system may accept a sentence syntactically yet still need domain definitions to interpret its content words appropriately. Versioning the rules matters when documents and tools are expected to agree.

Three practical examples

Three practical examples:
1. A documentation team uses approved equipment names and consistent instruction patterns. It tests whether readers can distinguish actions, conditions and warnings without relying on unexplained synonyms.
2. A multilingual publishing project limits avoidable ambiguity before translation. Translators still review meaning because constrained source text does not guarantee a perfect translation.
3. A knowledge-representation project writes statements in a language with defined formal semantics. The resulting logical interpretation is inspected before it is used for querying or consistency checks.

Think of it this way

Think of it: A controlled natural language resembles a carefully marked lane through a larger language. The lane offers fewer choices so that readers or software can follow agreed conventions. The analogy is limited because restriction has a cost: an author may need more sentences to express a nuance, and the lane's markings do not make every destination correct.

Worked example

Worked example: Define an original miniature instruction language with three rules: each sentence contains one imperative action; each action names its object explicitly; and object names must come from an approved list. The list contains “panel A” and “switch B,” and the permitted verbs are “open” and “close.” This is a teaching design, not a claim about any published standard.
The instruction “Open panel A and close it” violates the first rule by combining two actions and the second by using it rather than an explicit object name. If the intended second action concerns the panel, rewrite it as “Open panel A. Close panel A.” If it concerns the switch, write “Open panel A. Close switch B.”
The checker can identify the rule violations, but it cannot infer which second action the author intended merely from the ambiguous sentence. Nor do these rules establish that either action sequence is operationally appropriate. Language conformance, intended meaning and real-world correctness remain separate checks.

Illustrative case study

Illustrative case study: Fictional Pine Documentation maintains instructions written by several teams. Different authors use “cover,” “panel” and “door” for the same component, while “close it” sometimes refers to a nearby switch. The team defines preferred component names and requires explicit references in selected instruction types.
Editors trial the conventions on a small document set and ask representative readers to explain the intended actions. They revise rules that create awkward or misleading sentences. This invented case illustrates controlled authoring and user evaluation. It does not claim compliance with a named technical-writing standard or report measured improvements in safety, cost or translation quality.

Common misconceptions and mistakes

Common misconceptions and mistakes: A controlled natural language is not simply a list of keywords. Sentence construction and interpretation can matter as much as vocabulary. It is also not necessarily a programming language or a fully unambiguous logic notation. Different designs permit different levels of ambiguity and expressiveness. Passing a grammar checker does not prove factual accuracy. Finally, simplified wording can still hide an unclear condition, undefined term or incorrect assumption, so author training and review remain important.

Business use

Business use: Start with the communication problem that restrictions should solve. Define approved terminology and rules only as precisely as the workflow needs, then test author effort and reader understanding. Keep the rule version alongside checking tools and training material. Where a system translates controlled statements into executable queries or rules, show users the interpretation that will be used. Do not equate conformance scores with successful communication; a perfectly conforming statement may express the wrong requirement.

Check your understanding

Check your understanding: A document uses only approved words but combines them in a sentence that violates the chosen grammar. Is vocabulary compliance enough to establish language conformance?

Answer: No. The document must satisfy all relevant rules, which can include grammatical and interpretation requirements as well as vocabulary.

Related terms

Controlled vocabulary
Ambiguity
technical writing
Attempto Controlled English
formal semantics
requirements specification
predictive editor
natural language interface

Learn next

define a tiny controlled language for a harmless fictional task and write both accepted and rejected examples. Explain the meaning of each accepted sentence. Then identify one statement that follows the rules but is factually wrong, showing the limit of a conformance check.

Sources

Sources: Kuhn, A Survey and Classification of Controlled Natural Languages, 2014: https://aclanthology.org/J14-1005.pdf . Attempto, ACE 6.7 Construction Rules: https://attempto.ifi.uzh.ch/site/docs/ace_constructionrules.html ; ACE 6.6 Interpretation Rules: https://attempto.ifi.uzh.ch/site/docs/ace/6.6/ace_interpretationrules.html . The miniature instruction language and Pine Documentation case are original examples, not statements of ACE rules or compliance with another standard.

Previous: Controllability Next: Controlled vocabulary