Content Comprehension (Closed Beta)

Julia Randall
Julia Randall
  • Updated

Use the Content Comprehension feature to automatically understand assets, unlocking the full depth of content and creative metadata directly in The Data Standards Cloud.

Understand Content Comprehension

Content Comprehension uses computer vision to read the asset referenced in a template and then feeds rich AI-generated classification category data back into the template. AI-powered computer vision is superior to other contextual classification solutions because it organizes unstructured data such as images and videos, automates data labeling for assets at scale, and can process content in the absence of metadata. 

  • Set up Content Comprehension as part of Claravine's familiar template format, including the settings and dataset view with import/export capabilities you're used to.
  • Refine Content Comprehension with an adjustable confidence threshold. Each class/label combination has an associated score (ranging from 0-100). The higher the number, the better the prediction. Thus, you can select the strength of association and precision right for your classification needs.
  • Immediately visualize each asset in question with our image preview, alongside the image's metadata.

We currently support the following types of data analysis for image and video assets:

  • Visual classification: what is visually occurring
  • Text classification: metadata included as part of the asset processing
  • OCR: text that is present in the image or video
  • Audio: specific to videos and we can provide what language is present and/or a transcript

We can also support custom classes and labels to meet customer’s unique taxonomy needs.

Use Cases

With Content Comprehension, you can:

  • Extract a common taxonomy to better understand and classify videos and images, facilitating segmentation and analysis of your content and creative.
  • Connect classification metadata with campaign data for richer, more actionable measurement and insights.
  • Maximize the value of assets, campaigns, and data across the content supply chain.
  • Control the application of AI-generated metadata to assets in a DAM.
  • Preview an asset alongside its metadata to confirm creative placements.
  • Gain more, high-quality data for measurement and decisions.

Classification Categories

Classification categories include:

  • Logos
  • Places
  • Objects
  • Emotions
  • Activities
  • Affinities
  • Demographics
  • IAB Ad Product Taxonomy
  • IAB Taxonomy

Creative Asset Import and Export Capabilities

The following options are available to import creative asset URLs into a template:

  • Meta Ads Manager Inbound Integration
  • Adobe AEM Inbound Integration
  • Google CM360 Inbound Integration
  • Import files including publicly accessible creative asset URLs
  • Paste a creative asset URL into a manually created submission

For export, Content Classification AI fields can be delivered via our file-based outbound integrations, which include:

  • CSV export
  • Email
  • SFTP
  • AWS S3
  • Azure Blob Storage
  • Google Cloud Storage
  • Box

Limitations of Classification Data in Beta

It’s important to note that Classification data returned for creative assets is limited in compatibility with some platform features. Inbound and Outbound integrations are supported as noted above. Classification data is viewable in the Pending Submission and Dataset views.

Classification data is not compatible with template element features such as concatenation, split and extract, pick lists and powered lists, field sets, patterns, and integrations not listed under export capabilities.

Set Up Content Comprehension (Template Configuration)

Content Comprehension is set up during the template configuration process in The Data Standards Cloud. 

First ensure the following:

  • The template configuration must include fields for Creative Asset URL and Creative Name.
  • Creative asset URLs passed into The Data Standards Cloud must be publicly accessible.

To configure Content Comprehension in a template:

  1. Begin configuring a template and include fields for Creative Asset URL and Creative Name.

    CCConfig1.png
  2. Click Enable AI Creative Classification and follow the steps in the pop-up window.

    CCConfig2.png

  3. In the pop-up window, select the categories to include in the template.

    CCConfig3.png

  4. Select the boxes to indicate which template fields represent the Creative Asset URL and Creative Name.

    CCConfig4.png

  5. Adjust the Threshold for each Category to control the accuracy of data returned for each asset. Each class/label combination has an associated score (ranging from 0-100). The higher the number, the better the prediction.
    For example, a higher threshold means higher accuracy and likely a more narrow set of data returned. A lower threshold means looser parameters on classification accuracy and a larger breadth of data returned.

    CCConfig5.png

  6. Click Next to move on to configure inbound and outbound integrations.

Inbound Integration Configuration

Inbound options for creative assets include:

  • Meta Ads Manager, Adobe AEM, and Google CM360 inbound integrations
  • Data file import
  • Pasting a creative asset URL into a manually created submission

Meta Ads Manager Inbound Integration

You can use our inbound integration to import URLs for creative assets housed in Meta Ads Manager into our platform to be evaluated by Content Comprehension. 

To import Meta Ads Manager creative asset URLs:

  1. In the Meta Ads Manager Inbound configuration, under Data Selection, select Ad Creatives to ensure that creative data is imported from Meta Ads Manager. 

    MetaInbound1.png

  2. In the Field Mapping section of the Meta Ads Manager Inbound configuration, establish the required field mappings in order to import Meta creative fields.

    • Map the Meta Ad Creative Link URL field to your Claravine Creative Asset URL field.

    • Map the Meta Ad Creative Name field to your Claravine Creative Name field.

      MetaInbound2.png

Adobe AEM Inbound Integration

You can use our inbound integration to import paths for creative assets housed in Adobe AEM into our platform to be evaluated by Content Comprehension. Our AEM integration doesn’t use a creative asset URL. Instead, your AEM Auth credentials and the asset Path field are used to retrieve the asset each time Asset Image Preview and Content Classification are run. It’s typical to see a processing delay of a few seconds while we retrieve the asset for preview.

To import Adobe AEM creative asset paths:

  1. Rather than a Creative Asset URL field in the template configuration, the creative source will be the AEM Path.

    • Add a Path field to your template in the Define and Govern step.
      For Path, select the AI Classifications Source URL option.

    • Add a Name field which will map to the AEM Name field.
      For Name, select the AI Classifications Name option.

      AdobeInbound1.png

  2. In the AEM Inbound configuration, under Data Selection, select Assets to ensure that creative data is imported from Meta Ads Manager. 

    AdobeInbound2.png

  3. In the Field Mapping section of the AEM Inbound configuration, establish the required field mappings to import AEM creative fields.
    • Map the AEM Path field to your Claravine Path field.
    • Map the AEM cq:name field to your Claravine Name field. 

      AdobeInbound3.png

Google CM360 Inbound Integration

You can use our inbound integration to import URLs for creative assets housed in Google CM360 to be evaluated by Content Comprehension. We use the CM360 advertiser ID and creative name to automatically construct a URL (Generated Asset URL in the field mapping step below) to enable the creative asset to be evaluated.

To import CM360 creative asset URLs:

  1. In the CM360 Inbound configuration, under Data Selection, select Creatives to ensure creative data is imported from CM360.
    image.png
  2. In the Field Mapping section of the CM360 Inbound configuration, the required field mappings must be established to import CM360 creative fields. The  CM360 Generated Asset URL field must be mapped to your Claravine Creative Asset field and the CM360 Creative Name field must be mapped to your Claravine Creative Name field.
    image.png

Data File Import

You can import a CSV data file to populate the fields in your data set. Note that the columns and headers in the data file must match the configuration of the data set, and publicly available creative asset URLs must be included in the Creative Asset URL column.

DataFileImport1.png

Manual Submissions

You can enter creative data manually into a new submission. Note that publicly available creative asset URLs must be included in the Creative Asset URL column.

Outbound Integration Configuration

The Data Standards Cloud can deliver Content Comprehension AI fields via our file-based outbound integrations, including:

  • CSV export (using the Export Data button in the Dataset view to download a CSV file)
  • Email
  • SFTP
  • AWS S3
  • Azure Blob Storage
  • Google Cloud Storage
  • Box

Like you do with other Dataset fields, you can select the Content Comprehension categories as columns in the Outbound integration configuration. 


FileBased1.png

Content Comprehension in the Dataset

Content Comprehension begins evaluating creative assets and returning results as soon as the template is successfully created and creative data begins to flow into a Pending Submission. Clicking Save when finished configuring a template triggers Content Comprehension to run.

A note about timing: It takes time for Content Comprehension to run on creative data (several minutes for large datasets). In the Pending and Dataset views, Category data will be populated gradually.

Submissions

Content Comprehension data makes its way into the Dataset via a completed Submission. In Pending, even if comprehension hasn’t been completed for all rows of data, a Submission can still be completed. Completing the Submission triggers comprehension to run again on all rows of data included in the Submission to ensure up-to-date data in the Dataset.

In the Submission List view, the submission Status displays “Awaiting Classification” while the Content Comprehension processing is ongoing.

Submissions1.png

Once Content Comprehension is completed, the Status is “Success.”

Note: A new Submission is required to update comprehension data on existing rows in the Dataset.

In the Dataset

Once Submissions are complete, creative and classification data, along with a creative asset preview, become available in the Dataset. The Content Comprehension data columns are labeled for each Classification Category. They are displayed to the right of the creative columns and can include one or more values.

 

Dataset1.png

 

Comments

0 comments

Article is closed for comments.