SequenceMatch: Token-level string similarity comparison

This metric aims to evaluate token-level similarity between a generated label and a gold-standard label using Python's difflib.SequenceMatcher. It returns a similarity ratio in [0, 1] based on the longest contiguous matching subsequences shared by the two strings.

Example:

Generate_Label_1: hasVersion

Generate_Label_2: has_version

Generate_Label_3: Liquor

Gold_Standard_Label: hasVersion

Result:

1. SequenceMatch(Generate_Label_1, Gold_Standard_Label) = 1.00
2. SequenceMatch(Generate_Label_2, Gold_Standard_Label) ≈ 0.95
3. SequenceMatch(Generate_Label_3, Gold_Standard_Label) ≈ 0.20

What is being measured?

This metric captures structural overlap between two labels at the substring level. It is more forgiving than HardMatch and matches naming-convention variants, but it remains purely lexical and does not account for meaning.