Skip to main content

​
Input

  • output (str): The generated text to be evaluated.
  • expectedOutput (str): The reference or ground truth text.

​
Output

  • Result (float): A distance score between 0 and 1.

​
Interpretation

  • 0: The vectors have the same orientation (perfectly similar).
  • 1: The vectors are orthogonal (no similarity).
A lower distance corresponds to greater semantic similarity, with smaller values indicating that the two outputs are closer in meaning

​
Formula

Cosine Distance=1−A⋅B∥A∥×∥B∥\mathrm{Cosine\ Distance} = 1 - \frac{A \cdot B}{\|A\| \times \|B\|} Where A · B is the dot product of the vectors and ||A|| is the vector’s magnitude.
This is a distance metric distance = 1 - cosine_similarity. Lower scores (closer to 0) indicate higher similarity

​
How It Works

The evaluator computes embeddings for both texts and then calculates the angle between them. This provides a measure of similarity that is not affected by the length (magnitude) of the vectors.

​
Use Cases

  • Document similarity and text classification
  • Information retrieval and search engines
  • Paraphrase detection
⌘I