Skip to content
Nappai Documentation

Scorer

The Scorer component in Nappai helps you automatically rate items in your data. Think of it as a helpful assistant that grades your data based on the criteria you provide. It uses AI to understand your data and give a numerical score.

Relationship with AI Language Models

This component uses an AI language model (like those from OpenAI) to understand your data and criteria. The model analyzes your data and applies your rules to generate a score. The better the language model, the more accurate the scoring will be.

Inputs

  • Data: This is the information you want to score. It could be text, a document, or any data that the AI can understand. This is required.
  • Criteria: These are the rules the AI uses to score your data. Clearly describe what makes an item a high or low score. This is required. For example: “A high score means the text is well-written, grammatically correct, and factually accurate. A low score means the text is poorly written, contains grammatical errors, or is factually inaccurate.”
  • Additional Context: (Optional) Provide any extra information that might help the AI understand your data or criteria. This could be background information or specific instructions.
  • Max chunks: (Optional) This sets a limit on how many pieces the AI will break your data into for analysis. The default is 5. Only adjust this if you have very large pieces of data.
  • Include Justification: (Optional) If selected, the AI will explain why it gave a particular score. This can be helpful for understanding the scoring process.
  • Language Model: This is the specific AI model used for scoring. Nappai will usually select a suitable model for you.
  • Chunk size: (Advanced) This setting controls how large each piece of data is before the AI analyzes it. The default is 1500 characters. Leave this at the default unless you have a specific reason to change it.
  • Chunk overlap: (Advanced) This setting controls how much overlap there is between the pieces of data. The default is 150 characters. Leave this at the default unless you have a specific reason to change it.

Outputs

  • Score: A number between 0 and 100 representing the quality of your data based on your criteria.
  • Tool: (Technical output, usually not directly used) This provides a technical representation of the scoring process.

Usage Example

Imagine you want to score customer reviews.

  1. Data: You would input the text of the customer reviews.
  2. Criteria: You would provide criteria like “Positive language, mentions specific features, and overall satisfaction.”
  3. Run Scorer: Nappai will process the reviews and provide a score for each one. Reviews with positive language and mentions of specific features will receive higher scores.

Templates

[List of templates where the component is used will be populated here based on the system’s template configuration.]

[Links to related components with descriptions will be populated here based on the system’s component relationships.] For example, components that might be used to prepare data before scoring or to handle the results after scoring.

Tips and Best Practices

  • Clearly define your scoring criteria. The more precise your criteria, the more accurate the scores will be.
  • Start with smaller datasets to test your criteria and refine them before processing large amounts of data.
  • If you’re getting unexpected results, review your criteria and additional context to ensure they accurately reflect your needs.

Security Considerations

The data you input is processed by the AI model. Ensure that your data does not contain sensitive information unless the AI model and its environment are appropriately secured. Consult Nappai’s security documentation for more details.