Categorizer
Categorizer reads a text or a set of texts and tells you their main topics. With the help of an AI model, it proposes up to five categories, each with a name and a short description, sorted from the most to the least frequent.
It is useful when you have a lot of unclassified information and want to know what it is about before deciding what to do with it.
When to use it
- Discover the most common topics in customer reviews or surveys.
- Find the main reasons behind the tickets your support team receives.
- Group news or posts from your industry by topic.
- Give an agent a tool to classify the content it receives.
Configuration
| Field | What it does |
|---|---|
| Items | The texts or data you want to analyse. |
| Max Chunks | Maximum number of content fragments analysed. The text is split into fragments and only the first ones are reviewed. Default: 5. Increase it if the content is long and you want all of it considered. |
| Model | The AI model that identifies the categories. Connect a model component here. |
Output
- Categories: the list of detected categories (up to five), each with a name and description.
- Tool: lets an agent use the component as a tool and ask for the categories of a text whenever it needs them. When you use this output, you don’t fill in Items yourself: the agent sends the text.
Example
Your product team wants to know what users complain about most.
- Connect last week’s reviews to Items.
- Connect a model to Model and raise Max Chunks to
20so all reviews are covered. - Send the categories in a message to the team, for example: “Delivery times”, “Customer service” and “Packaging quality”.
Good practices
- If the result leaves out part of the content, increase Max Chunks. More fragments mean more time and more model usage.
- To assign each text to a specific category from your own list, use an agent and give it your list of categories instead.