Smart Content

Smart content makes it easy for creators and authors to discover content.

The Problem

Companies have thousands of digital assets (images, HTML pages, videos, PDFs, text, and such) in Oracle Content Management. These assets are valuable only when content creators, authors, and consumers can discover the content when they need it.

In order for content to be discovered, it must be tagged and classified. The content needs to be tagged with metadata describing the content and its meaning. Then the tagged content needs to be organized so that it can be found and delivered in the right context at the right time.

Up until now this tagging and classification has been a manual process. A median enterprise might spend hundreds of hours to tag and classify content, plus additional time to maintain accuracy. Failing to keep up with content tagging and classification leads to content discovery problems.

Oracle wanted to find a way to make content discoverable without the need for manual tagging and classification, and take it a step further and recommend the right assets without the user even needing to search.

The Solution

Smart content is Oracle Content Management's answer. We use artificial intelligence (AI) to analyze content, improving content discovery. Smart content includes the following features:

  • Smart search, which analyzes images for color and content and returns relevant images without the need for manual tagging. This improves the discoverability of images without having to painstakingly tag images.
  • Smart authoring, which recommends the right images during the authoring process. Users don't need to tag or search for images; images will be recommended based on the intent expressed in the current article.

    Note:

    If you're using Oracle Content Management Starter Edition, smart authoring isn't supported. To take advantage of the full feature set, upgrade to the Premium Edition.
  • Smart classification, which recommends categories based on the content in a content item.

Smart search with tag recommendations

How Smart Content Works

The backbone of smart content primarily consists of two types of AI models:

  • Computer vision models that are trained on millions of images to detect the relevant objects and parameters in new images. As new images are uploaded into repositories, they're automatically tagged to improve discoverability. When a user performs a search, the most relevant results are returned based on those automatic tags. Oracle Content Management also shows smart keywords (recommended keywords) to help users further refine their search.

    Note:

    You can still manually tag assets as needed to further improve the accuracy of search results.
  • Natural language processing (NLP) models that can derive the main intent from an article. This intent is used to recommend relevant images by matching the article intent to the automatic tags generated using the above computer vision models.

The Process

Smart content is enabled by default for any new repositories you create. For existing repositories, your repository administrator needs to manually enable smart content.

When a user searches for assets, Oracle Content Management will recommend search keywords.

When a user creates a content item that includes an image, Oracle Content Management will determine the intent of the article and suggest relevant images.

When a user creates a content item or assigns asset categories, Oracle Content Management will recommend categories based on the content in the content item.

Frequently Asked Questions

  • Is there a way to turn off smart content and undo any related changes?

    Yes. Turn off smart content in the repository.

  • How do I improve search behavior when smart search doesn't find the asset I'm looking for?

    If smart search doesn’t return the complete and accurate results for a search keyword, you can manually tag assets with the specific search terms you're looking for.

  • Why are smart content tags on images read-only?

    When you view the tags for an image, the tags shown under Smart Content Tags are the highest ranked machine tags. Machine tags are read-only because they aren't meant to be manually curated. Oracle Content Management analyzes images based on as many parameters and attributes as possible, leading to dozens of machine tags. In addition, our AI models are constantly improving, so machine tagging gets better as new images are added. If you're not able to find what you're looking for, you can always add manual tags. Oracle Content Management gives the highest importance to manual tags when returning results.

  • Does data leave Oracle servers?

    No.

  • Is any of our usage data being logged?

    Oracle Content Management does non anonymous logging to understand usage of the functionality in an effort to improve the service.