The Discover phase relies on unsupervised machine learning to examine the semantic similarity between messages and automatically group them into clusters representing latent themes in the data. Because no labels are required at this stage, users can explore these clusters to understand the natural structure and recurring topics within their communications, which then informs the definition of meaningful, business-relevant categories (labels) to be used when training a supervised classification model. This exploratory step improves the quality and relevance of the taxonomy created for subsequent supervised training.
Community Discussion