Knowledge Base

Introduction

ResourceSpace can use artificial intelligence to analyse images and text and generate useful metadata automatically.

AI can interpret the content of an image and produce information such as descriptions, keywords, titles and identifiable landmarks. It can also process existing text within a metadata field to extract key information, create summaries or categorise content.

This can reduce the amount of manual metadata entry required while improving the consistency, discoverability and organisation of digital assets. 

AI metadata generation in ResourceSpace is completely optional and, for our supported customers, runs locally so there’s no processing by a third party, no training on your data and no privacy concerns. All AI-generated information can (and should) always be reviewed by a human before being published.

Enhanced metadata creation

By analysing images and text, AI can generate structured and descriptive metadata based on the content of an asset.

For example, it can:

• Describe what is shown in an image

• Suggest relevant search keywords

• Generate titles and summaries

• Extract important information from existing text

• Assign assets to predefined categories

• Generate alt text automatically

Supporting rights management

AI can identify elements such as objects, scenes, visible text, logos and trademarks or people within images. This information may help users identify assets that require further rights or compliance checks.

***

Requirements

For supported customers AI functionality is included in your support contract and is ready to go. Contact us and we’ll get things set up to meet your requirements. Otherwise, the integration requires access to a supported AI service.

Depending on how the integration is deployed, you may need:

• An account with the selected AI service

• Suitable API credentials

• Sufficient usage allowance or credit

• Permission for ResourceSpace to send the selected image or text data to the service

Before enabling the integration, organisations should review the AI provider’s security, privacy, data-retention and processing terms.

Initial configuration

1. Navigate to Admin -> System -> Plugins.

2. Search for ‘AI Metadata Processing’ and select Activate.

3. Open the plugin options.

4. Enter the required connection details or API credentials.

5. Save the configuration.

Supported systems are already configured to use the AI service included with the selected ResourceSpace package.

Configuring metadata fields

The natural language instructions for the AI service are configured against the metadata field in which the response will be stored.

You must also select the source of the information to be processed. This can be either:

• The resource preview image; or

• Text stored in another metadata field

ResourceSpace uses the selected input together with the instructions and saves the response in the output field. The output will be updated whenever the value in the source field changes.

To configure a field:

1. Decide which metadata field should contain the generated output.
   Category tree and date fields cannot be used as output fields.

2. Open the metadata field configuration page.

3. Expand the Advanced options section.

4. Enter instructions in the AI prompt field.
   Write the instructions in natural language and be specific about the expected result. For example:

• “Create a concise description of this image.”

• “Return no more than ten keywords suitable for metadata indexing and searching.”

• “Summarise this text in no more than 50 words.”

• “Return the names of any countries mentioned, separated by commas.”

5. Under AI input field, select either:

• Image: Preview image; or

• The metadata field containing the source text

6. Select Save.

Using the integration

Once the fields have been configured, ResourceSpace processes the selected input and saves the generated response in the output field.

When the source is an image, the resource preview is analysed. When the source is another metadata field, the text from that field is processed.

Users should review generated metadata before relying on it for publication, rights decisions, sensitive classifications or automated workflows.

Possible uses

Automated image descriptions and tagging

ResourceSpace can generate descriptions and keywords based on the visual content of an image, improving searchability and reducing manual cataloguing.

Rights and compliance review

The integration can identify visible text, trademarks, logos, people and other potentially relevant content. This can help flag assets that may require further review.

Title and summary generation

Concise titles or summaries can be created from longer blocks of text, including:

• Document text

• Existing metadata

• Information imported from a collections management system

• Information imported from a product information management system

Keyword extraction

The integration can extract meaningful terms from text and return them in a format suitable for indexing and search.

Prompts can be made more specific by defining:

• The maximum number of keywords

• The required language

• The intended audience

• The desired level of detail

• The required output format

Automatic categorisation

AI can analyse asset content and suggest an appropriate category or classification. This can improve consistency and support reporting on the type and distribution of assets within the system.

Where categories must match an existing controlled list, the prompt should clearly specify the permitted values.

Processing existing resources

When the integration is enabled, existing resources may already contain suitable image or text inputs.

To retrospectively generate these values navigate to Admin->System->Jobs and in the ‘Trigger a job’ drop down, select ‘Process existing AI fields’ and ‘Configure job’. Select the AI field you’d like to populate. If you’d like to run this on selected collections only, enter the collection IDs in the field below and then choose if you’d like to overwrite existing values or not. Then click ‘Create job’. This will run as a scheduled task and can be monitored from the offline jobs page. 

Tips and troubleshooting

Usage monitoring and limits

For unsupported users who have configured a different model, system administrators can configure a usage limit and define the number of days over which usage is measured.

The plugin configuration page displays:

• The configured limit

• Current usage during the rolling period

• A warning when the limit has been exceeded

When the limit is reached, AI processing is paused until usage within the rolling period falls below the configured threshold.

These settings can be managed by a system administrator in the configuration.