Response Readability

Objective:The Response Readability metric measures the readability of a generated response. This score helps determine how easy or difficult it is for a human reader to understand the response produced by the language model.

Required Parameters: Response

Interpretation:

  • Higher Score: Indicates that the response is more difficult to read. This may be due to complex sentence structures, advanced vocabulary, or dense and convoluted text.

  • Lower Score: Suggests that the response is easier to read, with simpler sentence structures, common vocabulary, and clear, concise language.

Code Execution:

experiment_manager = Experiment(project_name="project_name",
                                experiment_name="experiment_name",
                                dataset_name="dataset_name")

response = experiment_manager.add_metrics(
    metrics=[
        {"name":"response_readability", "config": {"model": "gpt-4o"}},
        {"name":"response_readability", "config": {"model": "gpt-4"}},
        {"name":"response_readability", "config": {"model": "gpt-3.5-turbo"}}
    ]
)

Last updated