Concept clusters

Reporting > Text Processing > Topic & Theme Tagging > Concept Cluster

Note: This topic describes concepts and functionality for themes generated without artificial intelligence, also known as "legacy themes" and "themes 1.0."

This version of themes is still fully functional, and tag pools can still be configured to use legacy themes.

Themes with GenAI support all reporting modules and workflows supported by legacy themes. For more information, see Themes with generative AI.

The Concept Cluster screen displays the concept clusters generated by Medallia Text Analytics during the theme discovery process. You can add, edit, or delete concept clusters from this screen. For information about enabling theme discovery, see Configure Theme Explorer.
Note: This screen only appears after you select the Enable MATCH processing option (located at Reporting > Text Processing> Topic & Theme Tagging > Settings), click Save, and reload the page.

This topic includes information about these subjects:

Deleting and editing concept clusters during theme discovery (clustering)

You can opt to pause the theme discovery process to review the concept clusters before Text Analytics uses them to generate themes. Editing or deleting concept clusters during theme discovery impacts the themes generated for Theme Explorer once you resume the theme discovery process.

To delete or edit a concept cluster during theme discovery:

  1. On the Reporting > Text Processing> Topic & Theme Tagging > Topic Clusterings screen, select the Concept Editing Enabled property and then click Save.

    When Concept Editing Enabled is selected, the system pauses after creating the concept clusters, providing you with the opportunity to edit the concept clusters on the Concept Cluster screen.

  2. Run theme discovery (described in step 1 of Configure Theme Explorer).

    After the clustering process begins, you will receive an email when the concept clusters are ready for review.

  3. Navigate to the Concept Cluster screen and make the desired edits.

    Warning: One strength of the topic discovery engine is that it provides an unbiased view of the data. If you edit the clusters too much, you risk corrupting this unbiased view.
    • The concept cluster names become theme names. When a concept cluster name is misspelled, the capitalization is incorrect, or the name is otherwise unsuitable in some way, edit the name.

    • Do a spot check to verify that most of the words in the concept clusters seem appropriate. If there are a few anomalies, that is okay. Oftentimes. unexpected word associations reveal the most interesting insights.

      For example, in the following image, it is appropriate that the words budget, splurge, pricey, afford, and expensive are all members of the budget cluster, since those words are associated with the concept of a budget.

      The words in the Members field are associated with the selected concept.

    • When you want to prevent a concept cluster from becoming a theme, delete it. Unless the name of the theme is mentioned frequently, deleting the concept cluster will prevent a theme with the same name from appearing.

      For example, if you wanted to prevent a Pool theme from appearing, you would delete the Pool concept cluster:

      To delete a concept cluster, select the concept cluster and click Delete.

  4. Once you finish editing the concept clusters, click Resume processing for ‘tag pool’ and wait for theme discovery to finish.

Deleting and editing concept clusters after theme discovery (clustering)

After theme discovery has ended:

  • Editing the name of the concept cluster does not change the name of the corresponding theme.

  • Deleting a theme is not possible in Medallia Setup without running theme discovery (clustering) again.

  • Adding or removing concept cluster members only impacts how phrases from incoming survey records are matched to themes.

At this point, deleting its concept cluster only reduces the likelihood of the theme appearing in Theme Explorer. Phrases from incoming survey records are not matched to the theme unless the phrase includes the name of the theme. Since only the most popular themes appear in Theme Explorer, there's a chance that other, more-popular themes will take its place.

To delete a concept cluster, select the concept cluster and click Delete.

Properties

Name
Name of the concept cluster.
TagPool
Tag pool associated with the concept cluster.
Language
Language of the concept cluster.
Identifier
Concept cluster ID. Members of the concept cluster will map to this fixed identifier during theme discovery processing.
Members
Words in the concept cluster.