Multirate (Matrix Rating)

Name: Potato
Author: Potato

Configure a rating matrix in Potato where annotators score multiple items on the same scale simultaneously — useful for comparative evaluation and rubric-based tasks.

The multirate type displays multiple items in a matrix format where each item is rated on the same scale. Use it to evaluate several dimensions of a single item.

Multi-criteria rating interface Matrix rating interface for multi-dimensional evaluation in Potato

Basic Configuration

yaml

annotation_schemes:
  - name: "aspect_ratings"

Configuration Options

Field	Type	Required	Description
`name`	string	Yes	Unique identifier for the annotation
`description`	string	Yes	Instructions shown to annotators
`annotation_type`	string	Yes	Must be `"multirate"`
`labels`	array	Yes	Items to be rated (rows)
`options`	array	Yes	Rating scale options (columns)
`size`	number	No	Alternative to options: number of scale points
`min_label`	string	No	Label for lowest rating
`max_label`	string	No	Label for highest rating
`randomize`	boolean	No	Randomize item order
`compact`	boolean	No	Use compact layout

Examples

Response Quality Assessment

yaml

- name: "quality_assessment"

Translation Quality

yaml

- name: "translation_quality"

Product Review Dimensions

yaml

- name: "product_dimensions"

Output Format

The multirate annotation outputs a dictionary mapping each item to its rating:

json

{
  "id": "item_1",
  "annotations": {
    "aspect_ratings": {
      "Accuracy": "4",
      "Clarity": "5",
      "Helpfulness": "3"
    }
  }
}

Use Cases

LLM Evaluation: Rate responses on multiple quality dimensions
Translation Assessment: Evaluate fluency, adequacy, and terminology
Product Reviews: Capture ratings across different product aspects
Survey Research: Likert-style matrix questions
Peer Review: Rate papers on multiple criteria

Best Practices

Limit the number of items - 3-7 items works best; more causes fatigue
Use consistent scales - All items should use the same rating scale
Order items logically - Group related dimensions together
Provide clear definitions - Use tooltips to explain each dimension
Consider randomization - Prevents order bias in responses