Updated the spoken-word module prompts to preserve intentional emphatic repetition

src/audita/modules/prompts.py

@@ -102,6 +102,9 @@ def build_spoken_word_proposal_messages(section: TranscriptSection, glossary: Gl
         "Rules:\n"
         "- Approve only conservative cleanup of repeated words, repeated short phrases, filler words, hesitation artifacts, and similar spoken dysfluencies.\n"
         "- You may collapse adjacent repetition such as \"I I think\" to \"I think\" or remove filler spans such as \"you know\" or \"uh\" when local context supports that cleanup.\n"
+        "- Do not collapse repeated words or short phrases when the repetition plausibly expresses urgency, excitement, insistence, or deliberate rhetorical emphasis rather than dysfluency.\n"
+        "- Phrases such as \"Help! Help! Help!\", \"Stop! Stop! Stop!\", \"No! No! No!\", \"Yes! Yes! Yes!\", and \"Go! Go! Go!\" are often intentional emphasis and should usually be preserved.\n"
+        "- Only collapse repetition when local context supports it as accidental spoken repetition, hesitation, or verbal restart.\n"
         "- You may include low-risk punctuation, spacing, or capitalization cleanup when it is part of removing a dysfluency, such as removing ellipses or hesitation punctuation that no longer belongs after the cleanup.\n"
         "- Do not paraphrase, summarize, reorder ideas, replace content with different wording, or make substantive semantic edits.\n"
         "- Do not change clear content words just because a different phrasing reads better.\n"

src/audita/validators/prompts.py

@@ -69,6 +69,9 @@ def build_spoken_word_messages(validation_payload: List[dict]) -> List[Message]:
         "Rules:\n"
         "- Return one validation decision for every correction_index in the input.\n"
         "- Approve cleanup of repeated words, repeated short phrases, filler words, hesitation artifacts, and similar common spoken dysfluencies.\n"
+        "- Reject repetition cleanup when the repetition plausibly serves urgency, excitement, insistence, or deliberate rhetorical emphasis rather than dysfluency.\n"
+        "- Phrases such as \"Help! Help! Help!\", \"Stop! Stop! Stop!\", \"No! No! No!\", \"Yes! Yes! Yes!\", and \"Go! Go! Go!\" are often intentional emphasis and should usually be preserved.\n"
+        "- Approve repetition cleanup only when local context supports it as accidental repetition, hesitation, or verbal restart.\n"
         "- Approve minor punctuation, spacing, or capitalization cleanup only when it is plausibly part of removing a dysfluency.\n"
         "- Reject free-standing stylistic polishing, readability edits, paraphrases, and general rewriting.\n"
         "- Reject edits that materially change the segment's substantive meaning, even if they are not literal antonyms.\n"

tests/test_llm_validators.py

@@ -827,6 +827,8 @@ def test_spoken_word_prompt_emphasizes_dysfluency_cleanup():
 
     combined = messages[0]["content"] + messages[1]["content"]
     assert "dysfluencies" in combined
+    assert "intentional emphasis" in combined
+    assert "Stop! Stop! Stop!" in combined
     assert "punctuation" in combined
     assert "substantive meaning" in combined
     assert "original_segment_text" in messages[1]["content"]

tests/test_module_proposals.py

@@ -175,6 +175,8 @@ def test_spoken_word_prompt_is_explicitly_scoped_to_dysfluency_cleanup():
     combined = messages[0]["content"] + messages[1]["content"]
 
     assert "dysfluencies" in combined
+    assert "intentional emphasis" in combined
+    assert "Help! Help! Help!" in combined
     assert "punctuation" in combined
     assert "paraphrase" in combined
     assert '"id": 1' in messages[1]["content"]