Lexical Adherence

Lexical Adherence is Action Phrase's legacy Script detection algorithm. It listens to the spoken transcript and compares it against the text of each script section, looking for word overlap, phrase similarity, and section-by-section progress.

Instead of waiting for a specific cue phrase, Lexical Adherence tries to infer where you are in the script based on how closely your speech matches the written section.

Deprecated as of Action Phrase 1.1.0

Lexical Adherence is deprecated and kept for backward compatibility only.

Cue Points are replacing it because they are clearer, easier to edit, and more predictable in live use. With Cue Points, each section listens for a specific Magic Phrase instead of relying on inferred text overlap. This makes it much easier to understand why a section matched, adjust behavior, and avoid accidental jumps.

Future Removal

Lexical Adherence will be removed in a future version of Action Phrase. Existing scripts should be converted to Cue Points before then.

Use Convert to Cue Points from Script Options to keep your current sections and actions while generating short Magic Phrases from each section’s existing timing position. After conversion, review the generated Magic Phrases and add Alternate Phrases where needed.

How It Works

Lexical Adherence compares live speech against your script sections using deterministic text matching. It looks at section content, repeated word patterns, phrase overlap, and optional match hints to estimate the current section.

This can work well when you read closely from the script, but it becomes less predictable when you paraphrase, skip around, repeat earlier lines, or speak off-script.

Detection Timing

For Lexical Adherence scripts, each section can choose when its actions should run:

• Beginning: Run actions near the start of the detected section.
• Middle: Run actions around the middle of the section.
• End: Run actions near the end of the section.

This timing is estimated from the section’s text and the live transcript. Because it depends on matching progress through the section, timing may vary based on pacing, edits, skipped words, or improvised speech.