Configure Theme Explorer

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 first step in configuring Theme Explorer is to run theme discovery (clustering). During theme discovery, Medallia Text Analytics uses machine learning algorithms (rather than rule-based topics) to match phrases to topics. The themes created by the theme discovery engine appear in Topic Finder and Theme Explorer.

After theme discovery is complete, setting up Theme Explorer involves creating a new Theme Explorer standard report instance, adding the new report to a navigation subtab, and giving users permission to view the report.

Steps to configure Theme Explorer: 1. Run theme discovery 2. Add Theme Explorer.

For information about how customers can use the Topics/Themes Explorer dashboard module, see TA Topics/Themes Explorer. For more information about the legacy Theme Explorer report, see Theme Explorer standard report.

Restriction: The export feature that appears in Theme Explorer is not customizable.

There is an Export link in Theme Explorer.

Requirements

Theme Explorer requires a minimum of 1,000 comments. Medallia recommends a higher minimum of 5,000 comments, with at least 50,000 comments as ideal. Systems must also have multi-select filters enabled. Besides these requirements, the system requirements for the concept discovery engine also apply to Theme Explorer:

Before you begin

  1. Navigate to Reporting > Text Processing > Topic & Theme Tagging > Settings.

  2. If you plan to support sentiment, verify that TextAnalyticsVersion is set to SENTIMENT_SUPPORT. If it is not, follow the instructions outlined in Configuring sentiment.

  3. The name of the MainScoreField field appears in Theme Explorer, so users will see it. Navigate to Company > Profile > Company Settings and verify that name of the MainScoreField is appropriate.

    In this example, Likelihood to Recommend is in the MainScoreField.

  4. The name of tag pool appears in a Theme Explorer filter, so users will see it; navigate to Reporting > Text Processing > Topic & Theme Tagging > Auto Tagging Pools, and verify that the text in the Name field is appropriate.

  5. Turn on the Automatic Tagging Enabled property.

  6. If you want to exclude specific themes from clustering, select specific themes for each tag pool on the Reporting > Text Processing > Topic & Theme Tagging > Excluded Themes screen. For more information, see Excluded themes.

Running theme discovery

Running theme discovery involves clustering each phrase into one or more concept clusters, so it is a resource-intensive process. The time required for theme discovery varies depending on the number of phrases in the system and the number of systems running similar jobs.

Topic Finder appears automatically after you run theme discovery for the first time. No extra configuration steps are required. For Theme Explorer, however, your program must also have a Theme Explorer dashboard.

Important: Your Medallia Experience Cloud instance must have at least 1,000 records, else the theme discovery process fails.
Warning: Running theme discovery replaces any existing themes, and any previous edits made from the Topic clusterings for themes screen are lost.

Text Analytics clusters themes separately for each language. Text Analytics does not translate or merge themes because doing so might cause differences in vocabulary and nuances between languages to be missed. To view comment data in one group, spanning multiple languages, build topics. For more information, see Topics screen.

To run theme discovery:

  1. Verify that you have completed all the steps in the Before you begin section, including selecting the Enable MATCH processing property.

  2. Before you follow the next steps, verify that any processing job that was started recently has finished.

    If you go Reporting > Text Processing > Global Text Analytics Settings > Processings, the Current or last job status should be Completed.
    Warning: Do not start any clustering job before the processing job is complete.

    When the status says Completed, the processing job is complete. When the status is Running, the processing job is not complete.

  3. Navigate Reporting > Text Processing > Topic & Theme Tagging > Topic Clustering. See Topic clusterings for themes for details about this screen.
    Note: This screen appears after the Enable MATCH processing property is selected (see the Before you begin section).
  4. Select the Generate Word2Vec model checkbox.

  5. (Optional) If want to review and possibly edit the concept clusters, select the Concept Editing Enabled checkbox. After the clustering process begins, you will receive an email when the concept clusters are ready for review. See Deleting and editing concept clusters during theme discovery (clustering) for guidance on editing the concept clusters.

  6. If you want clusters to include terms introduced by user features, turn on the Theme Clustering Accounts for User Features property. For more information about user features, see User features.

  7. Select a tag pool from the Tag Pool dropdown.
    Tip: You can only select one tag pool at a time. If you need to run theme discovery on multiple tag pools, you'll need to complete this process for each tag pool.
  8. Click Save.

  9. In the Partial re-clustering frequency property, select how often you want Text Analytics to process incoming records for tag pools.

  10. Optionally, in the Email to property, enter email addresses (one per line) to receive clustering status updates.

  11. Select the Confirm checkbox and click Cluster topics for ‘tag_pool_name’.

    Once you change start the clustering job, the status listed in the Current to last topic clustering job field changes to Processing. The Processing status indicates that the job is actively running.

    In this example, the status says Processing, taken 864.00 ms so far.

    Warning: While the job is running, do not start or stop any processing jobs, backfills, or other clustering jobs.
    • If your tag pool has comments in multiple languages, Text Analytics creates themes in every language that has enough data.
    • Text Analytics does not use translated text to create themes. For example, when creating themes in English, Text Analytics only uses sentences that were originally written in English. However, once those English themes are created, they get applied to all English sentences (both comments that were originally in English, and comments that were translated from another language into English).

  12. (Optional) If you selected the Concept Editing Enabled property, select the Concept Clustering tab after you receive the email notifying you that the concept clusters are ready for review. See Deleting and editing concept clusters during theme discovery (clustering) for guidance on editing the concept clusters.

    After you finish editing the clusters, click Resume processing tag_pool_name and wait for the clustering process to finish.

After the theme discovery completes, you should not need to manually run the job again. Experience Cloud automatically runs a theme discovery job each month to update the available themes. This process occurs in the background, and does not require your involvement.

Note: After you complete these steps, themes appear automatically on the Topics screen. When you navigate to Reporting> Text Processing > Topic Sets and select the appropriate topic set, a list of themes appears. For more information, see Topic Finder.