ImageObject should support a dataMining field

[robrwo]

IPTC has added a new dataMining field for image EXIF data, see

• https://www.iptc.org/news/plus-publishes-draft-standard-for-image-data-mining-restrictions/
• https://www.iptc.org/std/photometadata/specification/IPTC-PhotoMetadata#data-mining
• http://ns.useplus.org/LDF/ldf-XMPSpecification#DataMining

It would make sense to add this attribute to the ImageObject or perhaps even the CreativeWork class.

[KalleOlaviNiemitalo]

Is the intention that a data miner would see this property in metadata, and then save time and I/O by not having to read the EXIF data?

[robrwo]

Is the intention that a data miner would see this property in metadata, and then save time and I/O by not having to read the EXIF data?

The intention is that somebody using this data would know that there are restrictions on how it can be used. Someone reading this data may not even have access to the image file, or the image format may not support EXIF data.

This data is used by search engines like Google (which uses it as an alternative to EXIF data when it's unavailable or fields are missing), so presumably Google would follow the restrictions. A common restriction would be "Prohibited except for search engine indexing".

A lot of organisations may have thousands or even millions of images online. Retroactively modifying all of these images to have a new EXIF field can be time consuming and expensive (costs of data transfers for cloud providers). It's can be much easier to update metadata for a web page, especially if it's generated from a database.

Also note that I am suggesting this attribute be added to the parent CreativeWork class. Certainly MediaObjects and DigitalDocuments would be appropriate.

[KalleOlaviNiemitalo]

http://schema.org/BusinessFunction seems to have prior art for how to import a third-party enumeration type to the http://schema.org/ namespace while referencing its original enumeration values.

[robrwo]

We could suggest the following values from http://ns.useplus.org/LDF/ldf-XMPSpecification#DataMining:

http://ns.useplus.org/ldf/vocab/DMI-UNSPECIFIED (Unspecified - no prohibition defined)
http://ns.useplus.org/ldf/vocab/DMI-ALLOWED (Allowed)
http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-AIMLTRAINING (Prohibited for AI/ML training)
http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-GENAIMLTRAINING (Prohibited for Generative AI/ML training)
http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-EXCEPTSEARCHENGINEINDEXING (Prohibited except for search engine indexing)
http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED (Prohibited)
http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-SEECONSTRAINT (Prohibited, see Other Constraints property)
http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-SEEEMBEDDEDRIGHTSEXPR (Prohibited, see Embedded Encoded Rights Expression property)
http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-SEELINKEDRIGHTSEXPR (Prohibited, see Linked Encoded Rights Expressio

[KalleOlaviNiemitalo]

How would "see Other Constraints property" etc. work here -- would the Other Constraints property have to be embedded in the file, or would that too correspond to some schema.org property?

[robrwo]

How would "see Other Constraints property" etc. work here -- would the Other Constraints property have to be embedded in the file, or would that too correspond to some schema.org property?

Cut and paste from the spec.

But then again, perhaps we should also add an "otherConstraints" field, similar to http://ns.useplus.org/LDF/ldf-XMPSpecification#OtherConstraints

It may also be worth "embeddedEncodedRightsExpression" and "linkedEncodedRightsExpression" if we want to implement all of these values.

[robrwo]

See #3392 which has some overlap

[github-actions]

This issue is being nudged due to inactivity.

[MatthiasWiesmann]

How much of this has been addressed in #3392.

[bquinn]

FYI in IPTC Photo Metadata we have two properties (among many obviously) which specify two different things:

• Digital Source Type specifies the source that caused this image to come into being. Was it taken by a camera? Scanned using a scanning device? Created as a screenshot on a computer or phone? Plotted using a mathematical formula (eg a fractal)? Drawn by someone using an Apple Pencil on an iPad? Created by a generative AI engine?
• Data Mining relates to the image/media owner's constraints about how this media object can be re-used as training data for AI engine to create generative AI or for other types of training (eg classification systems).

They cover two very different use cases.

Hope that helps to clarify.

[MatthiasWiesmann]

So would the proposal here be that a data-mining attribute be added which would echo the [digitalSourceType](https://schema.org/digitalSourceType) field? How much would be covered by a license field?

[robrwo]

So would the proposal here be that a data-mining attribute be added

Yes

which would echo the digitalSourceType field?

I don't know what you mean by "echo".

How much would be covered by a license field?

That depends. Is there a spec for the license field, and is it machine-readable and can express the same permissions/restrictions that the dataMining field can express?

Keep in mind that images and videos may already have the dataMining field in their metadata.