ODRL Vocabulary Document for the Unified Intent Mediator (UIM) Protocol Namespace

Date: September 23, 2024

Version: 0.1

Introduction

This document defines the Open Digital Rights Language (ODRL) Vocabulary for the Unified Intent Mediator (UIM) Protocol Namespace. The vocabulary is designed to address the need for a standardized taxonomy for AI data use, enabling website owners to express permissions, prohibitions, and obligations regarding the use of their data by AI systems.

By leveraging ODRL and defining specific terms within the UIM namespace, this vocabulary provides a framework for communicating data usage policies effectively and machine-readably, promoting responsible and ethical AI development.

Namespace Declarations

To ensure clarity and avoid naming conflicts, we declare the following namespaces:

{
  "@context": {
    "odrl": "http://www.w3.org/ns/odrl/2/",
    "uim": "http://uimprotocol.com/uim/odrl/",
    "skos": "http://www.w3.org/2004/02/skos/core#",
    "rdfs": "http://www.w3.org/2000/01/rdf-schema#",
    "xsd": "http://www.w3.org/2001/XMLSchema#"
  }
}

Vocabulary Definitions

This section defines the custom terms used in the UIM protocol namespace, including actions, constraints, duties, and assignee types. Each term is extensively documented to ensure clear understanding and interoperability.

Actions

uim:aiTraining

Type: odrl:Action
Label: "AI Training"
Definition: Use of data for training AI models.
Comment: Represents the use of data to train AI models, including machine learning algorithms and neural networks.

{
  "@context": {
    "odrl": "http://www.w3.org/ns/odrl/2/",
    "uim": "http://uimprotocol.com/uim/odrl/",
    "rdfs": "http://www.w3.org/2000/01/rdf-schema#"
  },
  "@id": "uim:aiTraining",
  "@type": "odrl:Action",
  "rdfs:label": "AI Training",
  "rdfs:comment": "Use of data for training AI models."
}

uim:aiResearch

Type: odrl:Action
Label: "AI Research"
Definition: Use of data for AI research purposes.
Comment: Covers the use of data in AI research, typically within academic or non-commercial settings.

{
  "@context": {
    "odrl": "http://www.w3.org/ns/odrl/2/",
    "uim": "http://uimprotocol.com/uim/odrl/",
    "rdfs": "http://www.w3.org/2000/01/rdf-schema#"
  },
  "@id": "uim:aiResearch",
  "@type": "odrl:Action",
  "rdfs:label": "AI Research",
  "rdfs:comment": "Use of data for AI research purposes."
}

uim:aiUse

Type: odrl:Action
Label: "AI Use"
Definition: General use of data in AI applications.
Comment: Encompasses any use of data in AI applications, including training, inference, and model development.

{
  "@context": {
    "odrl": "http://www.w3.org/ns/odrl/2/",
    "uim": "http://uimprotocol.com/uim/odrl/",
    "rdfs": "http://www.w3.org/2000/01/rdf-schema#"
  },
  "@id": "uim:aiUse",
  "@type": "odrl:Action",
  "rdfs:label": "AI Use",
  "rdfs:comment": "General use of data in AI applications."
}

uim:index

Type: odrl:Action
Label: "Index"
Definition: Crawling data for indexing and search purposes.
Equivalent to: odrl:reproduce
Comment: Used by search engines to crawl and index content for search purposes.

{
  "@context": {
    "odrl": "http://www.w3.org/ns/odrl/2/",
    "uim": "http://uimprotocol.com/uim/odrl/",
    "rdfs": "http://www.w3.org/2000/01/rdf-schema#",
    "skos": "http://www.w3.org/2004/02/skos/core#"
  },
  "@id": "uim:index",
  "@type": "odrl:Action",
  "rdfs:label": "Index",
  "rdfs:comment": "Crawling data for indexing and search purposes.",
  "skos:exactMatch": "odrl:reproduce"
}

uim:search

Type: odrl:Action
Label: "Search"
Definition: Making indexed data available for search queries.
Equivalent to: odrl:distribute
Comment: Involves providing search capabilities over indexed content.

{
  "@context": {
    "odrl": "http://www.w3.org/ns/odrl/2/",
    "uim": "http://uimprotocol.com/uim/odrl/",
    "rdfs": "http://www.w3.org/2000/01/rdf-schema#",
    "skos": "http://www.w3.org/2004/02/skos/core#"
  },
  "@id": "uim:search",
  "@type": "odrl:Action",
  "rdfs:label": "Search",
  "rdfs:comment": "Making indexed data available for search queries.",
  "skos:exactMatch": "odrl:distribute"
}

odrl:derive

We use the standard odrl:derive action to represent creating derivative works or AI models based on the data.

Constraints

odrl:purpose

Type: odrl:Constraint
Label: "Purpose"
Definition: Specifies the intended purpose of the action.
Comment: Used to constrain actions based on the purpose, such as non-commercial use.

uim:assigneeType

Type: odrl:LeftOperand
Label: "Assignee Type"
Definition: Specifies the type of assignee (e.g., non-profit, commercial).
Comment: Limits the policy to specific types of assignees.

{
  "@context": {
    "odrl": "http://www.w3.org/ns/odrl/2/",
    "uim": "http://uimprotocol.com/uim/odrl/",
    "rdfs": "http://www.w3.org/2000/01/rdf-schema#"
  },
  "@id": "uim:assigneeType",
  "@type": "odrl:LeftOperand",
  "rdfs:label": "Assignee Type",
  "rdfs:comment": "Specifies the type of assignee (e.g., non-profit, commercial)."
}

uim:dataFreshness

Type: odrl:LeftOperand
Label: "Data Freshness"
Definition: Specifies the allowed age of data for use.
Comment: Limits the use of data based on its age to ensure currency.

{
  "@context": {
    "odrl": "http://www.w3.org/ns/odrl/2/",
    "uim": "http://uimprotocol.com/uim/odrl/",
    "rdfs": "http://www.w3.org/2000/01/rdf-schema#"
  },
  "@id": "uim:dataFreshness",
  "@type": "odrl:LeftOperand",
  "rdfs:label": "Data Freshness",
  "rdfs:comment": "Specifies the allowed age of data for use."
}

uim:compensationRequired

Type: odrl:LeftOperand
Label: "Compensation Required"
Definition: Indicates that compensation is required for use.
Comment: Specifies that a fee or compensation is needed before the action is permitted.

{
  "@context": {
    "odrl": "http://www.w3.org/ns/odrl/2/",
    "uim": "http://uimprotocol.com/uim/odrl/",
    "rdfs": "http://www.w3.org/2000/01/rdf-schema#"
  },
  "@id": "uim:compensationRequired",
  "@type": "odrl:LeftOperand",
  "rdfs:label": "Compensation Required",
  "rdfs:comment": "Indicates that compensation is required for use."
}

Duties

odrl:attribution

Type: odrl:Action
Label: "Attribution"
Definition: Duty to attribute the source when using the data.
Comment: Requires users to provide appropriate credit to the content owner.

Assignee Types

uim:NonProfitOrganization

Type: skos:Concept
Label: "Non-Profit Organization"
Definition: Legally recognized non-profit organizations.
Comment: Organizations that operate for charitable, educational, or scientific purposes without profit.

{
  "@context": {
    "skos": "http://www.w3.org/2004/02/skos/core#",
    "uim": "http://uimprotocol.com/uim/odrl/",
    "rdfs": "http://www.w3.org/2000/01/rdf-schema#"
  },
  "@id": "uim:NonProfitOrganization",
  "@type": "skos:Concept",
  "rdfs:label": "Non-Profit Organization",
  "rdfs:comment": "Legally recognized non-profit organizations."
}

uim:CommercialEntity

Type: skos:Concept
Label: "Commercial Entity"
Definition: Entities operating for profit.
Comment: Businesses and organizations that engage in commercial activities.

{
  "@context": {
    "skos": "http://www.w3.org/2004/02/skos/core#",
    "uim": "http://uimprotocol.com/uim/odrl/",
    "rdfs": "http://www.w3.org/2000/01/rdf-schema#"
  },
  "@id": "uim:CommercialEntity",
  "@type": "skos:Concept",
  "rdfs:label": "Commercial Entity",
  "rdfs:comment": "Entities operating for profit."
}

JSON-LD Context

For completeness, the JSON-LD context that incorporates all the namespaces and prefixes used in the vocabulary is provided below:

{
  "@context": {
    "odrl": "http://www.w3.org/ns/odrl/2/",
    "uim": "http://uimprotocol.com/uim/odrl/",
    "skos": "http://www.w3.org/2004/02/skos/core#",
    "rdfs": "http://www.w3.org/2000/01/rdf-schema#",
    "xsd": "http://www.w3.org/2001/XMLSchema#"
  }
}

Policy Examples

Below are ODRL policy examples using JSON-LD serialization, demonstrating how the vocabulary can be applied to express various data usage policies.

1. Search Engines

Use Case: Search Engines (Allow Crawling Only for Indexing and Search Purposes)

Policy: Allow crawling only for indexing and search purposes.

{
  "@context": [
    "http://www.w3.org/ns/odrl.jsonld",
    {
      "uim": "http://uimprotocol.com/uim/odrl/"
    }
  ],
  "@type": "Set",
  "@id": "https://example.com/policy/search-engine",
  "permission": [
    {
      "target": "https://example.com/",
      "action": [
        {
          "id": "uim:index"
        },
        {
          "id": "uim:search"
        }
      ]
    }
  ],
  "prohibition": [
    {
      "target": "https://example.com/",
      "action": [
        {
          "id": "uim:aiTraining"
        },
        {
          "id": "uim:aiUse"
        }
      ]
    }
  ],
  "assigner": {
    "uid": "https://example.com/owner"
  }
}

Explanation:

Permission: Allows actions uim:index and uim:search on the target website.
Prohibition: Prohibits actions uim:aiTraining and uim:aiUse on the target.
Assigner: The policy is assigned by the website owner.

2. Non-Commercial AI

Use Case: Non-Commercial AI (Permit Data Use for AI Research or Development by Non-Profit Organizations or for Non-Commercial Applications)

Policy: Permit data use for AI research or development by non-profit organizations or for non-commercial applications.

{
  "@context": [
    "http://www.w3.org/ns/odrl.jsonld",
    {
      "uim": "http://uimprotocol.com/uim/odrl/"
    }
  ],
  "@type": "Set",
  "@id": "https://example.com/policy/non-commercial-ai",
  "permission": [
    {
      "target": "https://example.com/dataset",
      "action": {
        "id": "uim:aiResearch"
      },
      "constraint": [
        {
          "leftOperand": "purpose",
          "operator": "eq",
          "rightOperand": "http://www.w3.org/ns/odrl/2/non-commercial"
        }
      ],
      "assignee": {
        "uid": "*",
        "constraint": [
          {
            "leftOperand": "uim:assigneeType",
            "operator": "eq",
            "rightOperand": "uim:NonProfitOrganization"
          }
        ]
      }
    }
  ],
  "prohibition": [
    {
      "target": "https://example.com/dataset",
      "action": [
        {
          "id": "uim:aiTraining"
        },
        {
          "id": "commercialize"
        }
      ]
    }
  ],
  "assigner": {
    "uid": "https://example.com/owner"
  }
}

Explanation:

Permission: Allows uim:aiResearch action with constraints:
- Purpose Constraint: The purpose must be non-commercial.
- Assignee Constraint: Assignee must be of type uim:NonProfitOrganization.
Prohibition: Prohibits uim:aiTraining and commercialize actions.
Assigner: The policy is assigned by the dataset owner.

3. Attribution-Based AI

Use Case: Attribution-Based AI (Allow Data Use Only If the Resulting AI System Attributes Its Output Back to the Source Website)

Policy: Allow data use only if the resulting AI system attributes its output back to the source website.

{
  "@context": [
    "http://www.w3.org/ns/odrl.jsonld",
    {
      "uim": "http://uimprotocol.com/uim/odrl/"
    }
  ],
  "@type": "Set",
  "@id": "https://example.com/policy/attribution-ai",
  "permission": [
    {
      "target": "https://example.com/content",
      "action": {
        "id": "uim:aiUse"
      },
      "duty": [
        {
          "action": {
            "id": "attribute"
          },
          "target": "https://example.com/original-source"
        }
      ]
    }
  ],
  "assigner": {
    "uid": "https://example.com/owner"
  }
}

Explanation:

Permission: Allows uim:aiUse action with a duty to perform attribute action.
Duty: Requires attribution to https://example.com/original-source.
Assigner: The policy is assigned by the content owner.

4. Protecting Data Freshness

Use Case: Dynamic websites, such as news sites, limit the use of their content to prevent AI models from being trained on outdated information.

Policy: Limit the use of content to prevent AI models from being trained on outdated information.

{
  "@context": [
    "http://www.w3.org/ns/odrl.jsonld",
    {
      "uim": "http://uimprotocol.com/uim/odrl/",
      "xsd": "http://www.w3.org/2001/XMLSchema#"
    }
  ],
  "@type": "Set",
  "@id": "https://example.com/policy/data-freshness",
  "permission": [
    {
      "target": "https://example.com/news",
      "action": {
        "id": "uim:aiUse"
      },
      "constraint": [
        {
          "leftOperand": "uim:dataFreshness",
          "operator": "lt",
          "rightOperand": "P7D",
          "dataType": "xsd:duration"
        }
      ]
    }
  ],
  "assigner": {
    "uid": "https://example.com/owner"
  }
}

Explanation:

Permission: Allows uim:aiUse action with a constraint that data must be less than 7 days old (P7D).
Data Type: Specifies xsd:duration for the duration format.
Assigner: The policy is assigned by the news website owner.

5. Preventing Unwanted Commercial Use

Use Case: Websites behind paywalls or dependent on advertising revenue restrict the use of their data for commercial AI applications.

Policy: Restrict the use of data for commercial AI applications.

{
  "@context": [
    "http://www.w3.org/ns/odrl.jsonld",
    {
      "uim": "http://uimprotocol.com/uim/odrl/"
    }
  ],
  "@type": "Set",
  "@id": "https://example.com/policy/no-commercial-ai",
  "permission": [
    {
      "target": "https://example.com/premium-content",
      "action": {
        "id": "uim:aiUse"
      },
      "constraint": [
        {
          "leftOperand": "purpose",
          "operator": "eq",
          "rightOperand": "http://www.w3.org/ns/odrl/2/non-commercial"
        }
      ]
    }
  ],
  "prohibition": [
    {
      "target": "https://example.com/premium-content",
      "action": [
        {
          "id": "uim:aiTraining"
        },
        {
          "id": "commercialize"
        }
      ]
    }
  ],
  "assigner": {
    "uid": "https://example.com/owner"
  }
}

Explanation:

Permission: Allows uim:aiUse action for non-commercial purposes.
Prohibition: Prohibits uim:aiTraining and commercialize actions.
Assigner: The policy is assigned by the premium content provider.

Recommendations for implementers

Use Standard Terms Where Possible: To enhance interoperability, prefer standard ODRL terms when suitable. For example, use odrl:derive instead of defining a custom uim:derive action.
Define Custom Terms Carefully: When custom terms are necessary, ensure they are thoroughly documented and properly namespaced to avoid conflicts.
Ensure Policy Clarity: Policies should be unambiguous. Clearly specify permissions, prohibitions, and obligations to prevent misinterpretation.
Compliance with Legal Frameworks: Ensure that the policies and terms comply with relevant laws and regulations, such as GDPR for personal data protection.

References

ODRL Information Model: https://www.w3.org/TR/odrl-model/
ODRL Vocabulary & Expression: https://www.w3.org/TR/odrl-vocab/
ODRL JSON-LD Context: https://www.w3.org/ns/odrl.jsonld
SKOS (Simple Knowledge Organization System): https://www.w3.org/TR/skos-reference/
ISO 8601 Duration Format: https://en.wikipedia.org/wiki/ISO_8601#Durations
W3C RDF Schema (RDFS): https://www.w3.org/TR/rdf-schema/

Disclaimer

This document is provided as a template and requires adaptation to the specific needs and legal considerations of the implementing organization. It is recommended to consult legal experts when defining policies that have legal implications.

The UIM protocol authors and contributors are not responsible for any legal implications arising from the use of this document. However, we welcome feedback and contributions to improve the document and the UIM protocol.

We will strive to:

Maintain Updated Vocabulary: Regularly review and update the vocabulary to accommodate new use cases and changes in AI technology and data usage practices.
Promote Adoption: Encourage stakeholders to adopt this vocabulary to establish a common language for expressing AI data usage policies.
Provide Tooling Support: Develop tools and libraries to assist in creating, parsing, and enforcing ODRL policies using this vocabulary.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

uim-odrl-vocab.md

uim-odrl-vocab.md

ODRL Vocabulary Document for the Unified Intent Mediator (UIM) Protocol Namespace

Table of Contents

Introduction

Namespace Declarations

Vocabulary Definitions

Actions

uim:aiTraining

uim:aiResearch

uim:aiUse

uim:index

uim:search

odrl:derive

Constraints

odrl:purpose

uim:assigneeType

uim:dataFreshness

uim:compensationRequired

Duties

odrl:attribution

Assignee Types

uim:NonProfitOrganization

uim:CommercialEntity

JSON-LD Context

Policy Examples

1. Search Engines

2. Non-Commercial AI

3. Attribution-Based AI

4. Protecting Data Freshness

5. Preventing Unwanted Commercial Use

Recommendations for implementers

References

Disclaimer

Files

uim-odrl-vocab.md

Latest commit

History

uim-odrl-vocab.md

File metadata and controls

ODRL Vocabulary Document for the Unified Intent Mediator (UIM) Protocol Namespace

Table of Contents

Introduction

Namespace Declarations

Vocabulary Definitions

Actions

uim:aiTraining

uim:aiResearch

uim:aiUse

uim:index

uim:search

odrl:derive

Constraints

odrl:purpose

uim:assigneeType

uim:dataFreshness

uim:compensationRequired

Duties

odrl:attribution

Assignee Types

uim:NonProfitOrganization

uim:CommercialEntity

JSON-LD Context

Policy Examples

1. Search Engines

2. Non-Commercial AI

3. Attribution-Based AI

4. Protecting Data Freshness

5. Preventing Unwanted Commercial Use

Recommendations for implementers

References

Disclaimer