The Text Match metric is part of the Classic Metric Type options in Galtea. It leverages character-level fuzzy string comparison to assess whether two texts are similar enough, based on a predefined threshold. Rather than producing a continuous similarity score, this metric returns a boolean-style score, which is useful for scenarios where a definitive match decision is needed despite minor wording differences.

​
Evaluation Parameters

To compute the text_match metric, the following parameters must be provided:
  • actual_output: The generated text from the model.
  • expected_output: The target or reference text to compare against.

​
How Is It Calculated?

The metric uses a fuzzy string matching algorithm (based on edit distance) to compute the similarity ratio between the actual and expected outputs. If the similarity ratio exceeds 85%, the output is considered a match, otherwise; it is marked as a non-match.

​
Interpretation of Scores

  • 1.0 – Texts are considered a match (similarity > 85%).
  • 0.0 – Texts are not considered a match (similarity ≤ 85%).
This binary scoring makes the metric particularly useful for simple pass/fail validation use cases.

​
Suggested Test Case Types

Use Text Match when evaluating:
  • Simple equivalence checks where light paraphrasing or minor differences are allowed.
  • Rule-based or heuristic outputs where exact matches aren’t expected but alignment is necessary.
  • Pass/fail QA checks for generated text.