User permissions required: ‘View Sources’ AND ‘View Labels’
The Discover page performs two key functions:
1. Firstly, it helps you to discover interesting clusters of verbatims. Clusters are themes of verbatims, which the platform has identified as sharing similar intents or concepts.
When data is uploaded, the platform uses unsupervised learning (i.e. it reads and interprets the data without any human training) to automatically discover these clusters of similar verbatims and present them in the GUI. This functionality makes discovering new intents and applying labels easier and faster and is typically the first step in the model training process.
After a significant amount of training is completed, or a significant amount of data added to the dataset, Discover will retrain and present you with new clusters of unreviewed verbatims. When Discover retrains, it takes into account the existing taxonomy, in order to present you with new clusters that are still interesting to you.
2. Secondly, Discover allows users to label verbatims in bulk, as well as individually, using either the 'Cluster' function (discussed above), or the 'Search' function. As the verbatims in each cluster should contain similar intents and concepts, the bulk label functionality is a helpful tool to quickly train the model.
The search function allows you to search for key terms or phrases that you know may be relevant for certain labels that you want to capture.
To understand more about how to use Discover, see here.