Topic clusterings for themes
Reporting > Text Processing > Topic & Theme Tagging > Topic Clusterings
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 time required for theme discovery varies based on the size and number of jobs on the processing queue, and the volume of phrases in your system.
The Topic Clusterings screen appears after you turn on the Enable MATCH processing property on the Reporting > Text Processing > Topic & Theme Tagging > Settings screen, click Save, and then refresh your browser screen.
Properties
- Current or last topic clustering job
- Details about the current or more recent theme discovery job.
- Theme Clustering Accounts for User Features
- When on, the theme clustering process accounts for user features (which otherwise might not be considered as words, or be counted as separate words). This feature requires tokenization, so ensure that the Skip Tokenization property is off when processing is required. For more information about user features, see User features.
When using this feature, consider also using excluded themes, as described in Excluded themes.
- Skip Tokenization
- Whether or not to skip the first step of the theme discovery process, which is called tokenization. The records in the system only need to be tokenized once.
After you select the Enable MATCH processing property on the Reporting > Text Processing > Topic & Theme Tagging > Settings screen, all your records are tokenized the next time you initiate text processing (parse your comments). Tokenizing during theme discovery is usually unnecessary because in most cases, the records have already been tokenized.
When on, Experience Cloud skips the tokenization step. Tokenizing is usually unnecessary. However, if you skip this step and the records have not yet been tokenized, theme discovery cannot progress. The process will fail quickly and the system will display the following message. If this happens, deselect this option, save the page, and start the theme discovery progress again.
When off, Experience Cloud tokenizes the records.
Note: Tokenizing is a time-intensive process; in some cases, this step takes more than a week. If you think Experience Cloud needs to tokenize records, contact Medallia Support before turning this property off. - Generate Word2Vec model
- Whether or not processing generates the Word2Vec model automatically. Generating this model is an intermediate step of the theme discovery process.
When on, Experience Cloud does not pause before generating the Word2Vec model. When off, Experience Cloud does pause.
Warning: Do not turn this property off unless you are a member of the Medallia Engineering team. - Concept Editing Enabled
- Whether or not the process pauses after the concept clustering phase, allowing you time to edit the concept clusters. Select this property only if you are a member of a Medallia implementation team and want to review and edit the concept clusters.
After you finish editing concept clusters, click Resume processing for ‘tag pool’ and wait for theme discovery to finish.
Important: If Concept Editing Enabled is turned on, you cannot complete a partial recluster. Concept Editing Enabled must be off to run a partial recluster. - Threshold for theme activation on partial publish
- Period of time after which themes are activated on a theme publish flow, making them available in reporting. This enables activation of themes before the entire publish job is finished.
- Partial re-clustering frequency
- Select a frequency at which topics are reclustered, from Monthly (1st day of every month), Twice a Month (1st and 15th day of every month), and Once a Week (Monday of every week). By default, topics are reclustered monthly.
If a tag pool has already been clustered, a button appears that enables you to manually start a partial reclusting of the most recent records (by response date) for the span of time configured by this property. For example, if you have the Partial re-clustering frequency property set to Twice a Month (1st and 15th day of every month) for the Retail Store tag pool, the button is Trigger a Partial Recluster on 'Retail Store' for the last 15 days.
Important: If Concept Editing Enabled is turned on, you cannot complete a partial recluster. Concept Editing Enabled must be off to run a partial recluster. - TagPool
- Name of the tag pool to process. You can only process one tag pool at a time. Warning: After changing this field, refresh the button by saving and reloading the page. After you select this option for the first time, a button appears that allows you to cluster, recluster, resume, or cancel the theme discovery process. The content of this field affects the function of the button underneath it, so always save and refresh the page after saving this option.
- Email to
- A list of email addresses (one per line) that receive clustering process status updates.
