trustMinder allows verbatim text to have
- Sentiment Analysis (NLP)
- Single label classification based on a ML model
- Multi-label classification based on a ML model
This solution covers multi-label classification.
A multi-label classification model requires significant training data but once available it can return the manual time & effort required to code new verbatims. The ML model will use a set of defined labels.
The CustomMLClassify feature is only available if a corresponding Machine Learning (ML) Model exists.
The ML will apply the defined labels to the verbatim if the score associated with the label achieves a threshold level.
The threshold is determined by the client but typically is within the range of 0.6-0.9
This means a piece of text may have several labels and is influenced by the threshold level. The threshold is set by trustMinder support - it is not available to set in your trustMinder account.
In the Insights Manager section, to view the multi-label classification in a bar chart you should:
- Select the verbatim question from the Metric drop-down list
- Select CustomMLClassify from the Feature list
- Select Bar from the Chart type list
- Click Run Report
Please note the entire list of defined labels will appear on the chart. This allows for similar charts to be quickly compared.
To view a trend of the multi-label classification model
- Select the question from the Metric drop-down list
- Select Month from Group
- Select CustomMLClassify from the Feature list
- Select Line from the Chart type list
- Click Run Report