Use Automatic Relevance Tuning

In this article

Automatic Relevance Tuning (ART) will improve the base relevance of your search experience based on user behavior (see About Automatic Relevance Tuning).

Like any other Coveo Machine Learning (Coveo ML) model, an ART model must be associated with a query pipeline to be leveraged. Hence, enabling Automatic Relevance Tuning begins with performing the following steps in Coveo:

  1. Creating a query pipeline.

  2. Creating an Automatic Relevance Tuning model. As part of the ART model creation process, you will associate it with your query pipeline.

To use the output of your ART Coveo Machine Learning model, a query must be routed to the corresponding query pipeline. This is achieved by configuring a Coveo Hive rendering to use the desired query pipeline. The following Coveo Hive renderings include a data source option in which you can specify which query pipeline to target:

Troubleshooting

Automatic Relevance Tuning issues you may encounter are typically a result of the usage analytics data your ART model has learned from (or lack thereof). For more details on how to train your ART model learning and review its learning dataset, see Troubleshoot ART models.

Note

The Troubleshoot Automatic Relevance Tuning models article mentions that the contentIdKey and contentIdValue parameters must be present in the customData of click events. Don’t worry about this. Coveo for Sitecore renderings automatically add these parameters to the click event calls.