textSimilarity

function textSimilarity(a, b): number;

Defined in: src/utils/textSimilarity.ts:42

Compute a similarity score (0–1) between two transcript strings.

Parameters

Returns

number

A similarity score from 0 (completely different) to 1 (identical).

Remarks

Uses an order-aware word overlap approach: normalizes both strings to lowercase, splits into words, then checks how many words from the shorter string appear — in order — in the longer string. This handles the most common eager-pipeline scenario where EndOfTurn text is a superset of EagerEndOfTurn text (e.g., “hello world” → “hello world how are you”).

• Returns 1.0 when the texts are identical (after normalization).
• Returns 1.0 when the eager text is a perfect prefix of the final text.
• Returns a value between 0 and 1 based on the proportion of matching words when the texts diverge.
• Returns 0 when the texts share no words.

Example

textSimilarity('hello world', 'hello world');           // 1.0
textSimilarity('hello world', 'hello world how are you'); // 1.0 (prefix match)
textSimilarity('hello world', 'goodbye world');         // 0.5
textSimilarity('hello', 'goodbye');                     // 0.0