Skip to main content
Helix separates classification into four dimensions. Each dimension asks a different question about the same analysis unit.

Why the dimensions stay separate

An email address is an Entity. An email message is a Document. A discussion of cybersecurity is a Topic. A statement that an employee works for a company can express a Relation. Collapsing these meanings into one label list would create ambiguous results. Keeping them distinct lets each task use the prediction unit and geometry appropriate to its question while preserving one coherent semantic system.

Hierarchy preserves useful precision

Concepts can progress from broad to specific. Evidence should resolve to the deepest meaning it supports, while broader ancestors remain available for navigation and interpretation. A system must not invent a more specific child when the evidence supports only the parent.

Meaning is not availability

A semantic concept can exist before a particular Helix product surface supports detecting it. Public capability claims therefore require separate qualification; they are never inferred from the breadth of the underlying specification.
Classification describes what the available evidence supports. Customer policy determines whether that meaning is sensitive for a particular use and what action, if any, should follow.