Multi-Class and Multi-Label Text Tagging in Label Genie
- Predictive AI
- H2O Cascade
This lesson extends single-class labeling to a multi-class, multi-label setup in H2O Label Genie, using product reviews that can express sentiment about more than one product at once. It walks through creating a new classification task on the Amazon Reviews demo dataset, defining four classes such as satisfied or dissatisfied with each of two products, and enabling the option to allow multiple labels per entry. Once zero-shot predictions finish running, it shows how some reviews end up with more than one label checked, reflecting comments that mention two products or mix more than one sentiment in the same entry.
Key takeaways
- How to define multiple classes, like satisfied or dissatisfied with two different products, in one task
- Why enabling multi-label selection matters when a single entry can express more than one sentiment
- How zero-shot predictions can flag more than one label on a review that discusses multiple products
- What the prediction percentage shown for each class tells you about the model's confidence
Start owning your AI