Models
A Model is the AI that reads every piece of customer feedback and labels it. Classifiers tag feedback with labels, such as the topics customers talk about (content classifiers) or how they feel (sentiment classifiers); Dimension models answer a research question, such as who is talking or what goes wrong, for the Dimensions; extractors pick out entities. A model can be multi-label, giving one piece of feedback several labels, or single-label, giving it one.
Kimola has 15 pretrained models ready to use, from the Automatic Classifier, which discovers themes from your data, and the Sentiment Classifier to industry classifiers for e-commerce, mobile apps, software, banking, automotive and more. Pretrained models work in every package. Create a custom model when you want feedback labeled with your own themes, the same way every time: you can build one from a prompt, from a training dataset or from a Report you already have. Custom models are in the paid packages, not in Starter.
Every analysis in Kimola starts with models. A Report classifies its feedback with the models you choose, and a new Report uses the Automatic Classifier and the Sentiment Classifier by default. A Funnel, in the Enterprise package, combines your custom models so every Space labels its feedback the same way.
Every model an analysis uses counts in its queries, by the rule rows × models × Dimensions explained in Account. Custom models can be grouped with related Reports and Feeds in a Project.
- What is a Model?
- Custom vs. Pretrained Models
- What is a Custom Model?
- Create a Custom Model with a Prompt
- Create a Custom Model with a Training Dataset
- Optimize Dataset-Based Custom Models
- Create a Custom Model with a Report
- Organize Labels into Categories
- What is Model Consistency?
- Explore a Custom Model
- Manage Custom Models
- Test a Custom Model
- What is a Pretrained Model?
- List Pretrained Models
- Test a Pretrained Model
- Multi-label vs. Single-label Classification