As content authors build more and more content managing and surfacing that content becomes a bigger challenge. A common request is how can content authors tag their content with relative information to help organize and surface it. With Sitecore, this can be done via Sitecore Cortex Content Tagging.
As with almost everything in Sitecore it is an extensible framework. It uses the Open Calais AI to process content and return standard tagging information. To enable this there is really not much that is needed. The Sitecore documentation has most of the steps you need and there are a few other blog posts out there with details as well. I found what all these are lacking though is a working sample of the configuration you need. All you need is a patch file like this (just add your API token):
What can be a little tricky is where to find your API Key. Just go to https://permid.org/ login and then in navigation click on APIs and the big button that says "Display My API Token".
Once you have set this up and followed steps outlined in the documentation to tag some content how do you customize and manage this process. First, let's start with where do all these tags go? After the API call is made to Open Calais Sitecore pushes that tags back into Sitecore and stores them here /sitecore/system/Settings/Buckets/TagRepository. This is a Sitecore Item Bucket so you can search for certain tags and it can grow large over time and still be manageable.
Now let's look at what happens when running this process and applying tags.
Architecture:
- Providers: Provide the essential business logic for how content is tagged.
- IContentProvider: Extracts the content from Sitecore and preps it for analysis.
- IDiscoveryProvider: Processes the taggable content and extracts tag data
- ITaxonomyProiver: Tags all discovered tag data and creates a list of tags that should be assigned.
- ITagger: Assigns tags to the object being tagged.
- Configuration: Control how all the business logic is pulled together.
- Pipelines: Extension points for granular control over tagging
- getTaggingConfiguration: This determines the configuration set to use. It holds two processors.
- GetDefaultConfigurationName: This pipeline is pretty simple and just looks for a setting with the key of Sitecore.ContentTagging.DefaultConfigurationName and returns its value (surprise it is "Default")
- BuildConfiguration: This will use the select configruation name to build and return an ItemContentTaggingProvidersSet.
- tagContent: Run configured providers
- RetriveContent: Uses the configured IContentProviders to select fields. The default provider will not process anything that starts with a double underscore "__" and only fields and types defined in the configuration.
- Normailize: This process is actually responsible for trigging another pipeline of "normalizeContent" which out of the box only has one process that stripes HTML from the field.
- GetTags: This is going to loop through all DiscoveryProviders to get tags for the content generated in steps above. The by default is the OpenCalaisProvider. If you want to get a feel for what content is returned from Open Calais you can use their test API. For Open Calais 4 groups are returned. This provider imports all of these groups regardless of their relevance score.
- SocialTags: Social tags are derived from the Wikipedia folksonomy. They are periodically updated to keep them current.
- Industry: Industry tells you the industry it thinks the content aligns to.
- Entities: This can have references to companies and products for those companies.
- Topics: What is the general topic of the content.
- StoreTags: This will take all the tags that were returned and add those tags to the above-mentioned storage location.
- ApplyTags: This will save the needed tags to the "selected" area for the item that was being tagged.
- normalizeContent: Prepares TaggableContent
The Sitecore.ContentTagging.Core.config sets up default providers for all of these. Note that there is also a Sitecore.ContentTagging.OpenCalais.config that overrides the default provider with the Open Calais provider.
Points to Note:
- Open Calais is free but it does have limits. The main limit being it will only process 1,000 records a day.
- Using Open Calais means using a standard ontology. If you need a custom ontology or other customizations you will need a different provider or something in the pipeline that applies your customizations.
Comments