AI Glossary / Data engineering and architecture / Advanced
Metadata object
Definition
A metadata object is an object type defined within a metamodel for describing metadata. This is the technical meaning used in the OECD glossary and ISO/IEC 11179 terminology. An individual instance of such a type is a metadata item. The distinction matters because a model can define one kind of descriptive object while a registry contains many individual descriptions that instantiate it.
What it means
The term refers to part of the structure used to organize descriptions, rather than automatically meaning a particular document or dataset. In an original learning-catalog model, a type called DatasetDescription could establish the properties used to describe datasets. Separate descriptions of course enrollments and resource usage would then be instances of that type. The type helps make those descriptions comparable, but it does not make the datasets themselves identical. This modeling sense should also be distinguished from informal software usage in which developers call almost any metadata-bearing record an object.
How it works
A metamodel defines relevant types and the relationships or attributes associated with them. Metadata objects occupy that type level, while metadata items instantiate them. ISO's public terminology preview and the OECD glossary establish this distinction. In an original local design, a DatasetDescription type could have a definition attribute and a relationship to a responsible-party description. The designer would specify allowed values and relationships as needed. These illustrative choices are not a reproduction of the ISO metamodel. A production implementation must identify the actual model and version it follows before claiming that its structure conforms to a standard.
Three practical examples
- A metadata registry defines a common type for descriptions of datasets contributed by different teams.
- An AI catalog project distinguishes the template-level structure for data-element descriptions from the individual descriptions it stores.
- A migration review checks whether two systems model the same kind of metadata object before mapping their individual records.
Think of it this way
Imagine a library deciding what kinds of catalog records it needs, such as book descriptions and publisher descriptions. Each kind has its own expected properties and relationships. The kind of record is different from the completed description of one particular book. The analogy illustrates the model's types and their instances. It does not assert that every library catalog implements the formal metadata terminology in exactly the same way.
Worked example
Consider an original simplified metamodel with two metadata object types: DatasetDescription and FieldDescription. DatasetDescription has a name and purpose. FieldDescription has a name, definition and unit note, plus a relationship to a DatasetDescription item. A small registry contains one dataset-description item D1 for Resource Measurements and three field-description items F1, F2 and F3 for resource identifier, duration in seconds and publication date. There are two metadata object types and four metadata items. The three field items each relate to D1, producing three field-to-dataset relationships. Adding a fourth field-description item F4 for review date increases the item count to five while leaving the type count at two. By contrast, adding a new metadata object type called TransformationDescription changes the metamodel itself. The example's counts assume relationships and attributes are modeled as specified here; richer models may represent them using additional types. It is a teaching model rather than a normative registry design.
Illustrative case study
In a fictional AI data-catalog project, a team describes every newly imported field as a new metadata object type. The model quickly becomes difficult to understand because ordinary records are mixed with structural definitions. Reviewers separate the two-level discussion: FieldDescription is a type, while the individual duration and date descriptions are items. They then identify a genuinely different requirement to describe transformations between measurements and display values. That requirement may justify extending the model rather than simply adding another field item. The scenario illustrates choosing the correct modeling level. It does not claim that the proposed two-type design is sufficient for a complete production registry.
Common misconceptions and mistakes
A metadata object, in this technical usage, is not simply any file that contains metadata. Another mistake is assuming that adding an item always changes the metamodel. An existing type can support many additional instances without altering its definition. The reverse error is treating a structural change as an ordinary content edit, even when it affects how many items are interpreted. Product terminology may also differ from standards terminology. Readers should check the applicable model rather than infer meaning from the word object alone. Finally, a diagram's neat appearance does not prove interoperability or conformance.
Business use
For an AI-assisted catalog, establish a shared vocabulary for model changes and record changes. In the fictional registry, adding a review-date description is different from introducing a new type for transformation descriptions. Assign appropriate review to each kind of change and inspect how existing items would be affected by revised type definitions. When exchanging metadata with another system, compare the meaning of the types and relationships before mapping field names. This helps teams avoid confusing ordinary catalog growth with a structural redesign and makes the scope of migrations easier to explain.
People also ask
Is a metadata object the same as a metadata item?
In the terminology used here, no. The object is a type defined by a metamodel, and the item is an instance.
Does adding another field description necessarily add another type?
No. It may instantiate an existing type.
Do all software products use object this way?
No. Informal and product-specific terminology can differ, so the governing model should be identified.
Check your understanding
How many metadata object types are in the initial toy metamodel?
Answer: Two.
How many items are in its initial registry?
Answer: Four.
What happens when F4 is added?
Answer: The item count becomes five, but the type count remains two.
What changes when TransformationDescription is introduced as a new type?
Answer: The metamodel's structure.
Why are the counts explicitly scoped to the toy model?
Answer: Other models can represent attributes and relationships differently.
Related terms
- Metadata item
- Metadata registry
- Data model
- Metamodel
- object type
- schema evolution
Learn next
Sketch a metamodel with two types and a relationship between them. Add three items without changing the model, then propose one requirement that genuinely needs a new type or relationship. Explain which existing items would require review and why the change is structural rather than merely another record.
More in Data engineering and architecture
- Metaclass (knowledge representation) - In knowledge representation, a metaclass is a class whose members are themselves classes.
- Metadata content standard - A metadata content standard specifies which descriptive information is to be provided about data or other resources.
- Metadata controller - A metadata controller, or MDC, coordinates file-system information and access in certain shared-storage architectures.
- Metadata repository - A metadata repository is an organized store of information describing data, systems and their relationships.
- Metadata structure definition - A metadata structure definition, commonly abbreviated MSD in SDMX, specifies the organization and allowed content of a reference metadata set.
- Metadatabase - A metadatabase is a database organized around metadata: descriptions of other data, models and relationships.
Sources
OECD, Glossary of Statistical Terms, 2008, Metadata object: https://www.oecd.org/content/dam/oecd/en/publications/reports/2008/09/oecd-glossary-of-statistical-terms_g1gh9ad7/9789264055087-en.pdf ISO Online Browsing Platform, ISO/IEC DIS 11179-5 public terminology preview, citing ISO/IEC 11179-3:2023: https://www.iso.org/obp/ui?_escaped_fragment_=iso:std:iso-iec:11179:-5:dis:ed-4:v1:en The two-type model, counts, library analogy and AI catalogue scenario are original educational illustrations. The public ISO draft preview supports terminology; no full-standard conformance assessment or universal product terminology is claimed.
Previous: Metadata item Next: Metadata registry