Audit malformed Ego4D voice segments

docs/ego4d_av_llm_annotation_review.md

@@ -55,6 +55,18 @@ handoffs are added as independently reviewable hard-negative proposals.
 Timing rules only select and describe candidates; they never assign semantic
 positive labels.
 
+Official zero-duration voice records cannot define an activity interval. They
+are never expanded to an invented duration: the generator skips them, reports
+the count on stdout, and stores their complete source records under
+`clip_audits` and each candidate's `temporal_evidence`. A negative-duration
+record remains a hard error.
+
+Ego4D v1 also contains nine train clips whose official voice endpoint extends
+past `clip_end_sec` by 0.42--1.03 seconds. The default audited ceiling is 1.1
+seconds. These endpoints are preserved and flagged rather than silently
+clamped; larger overruns remain errors. A human should normally exclude a
+semantic boundary beyond observable media as `DATA_RECORDING_BOUNDARY`.
+
 ## 2. Inspect what will be uploaded
 
 The LLM client refuses to upload transcript excerpts without an explicit

prepare_ego4d_av_eot_int_review.py

@@ -28,7 +28,7 @@ class CandidateConfig:
     relation_window_seconds: float = 3.0
     same_speaker_pause_max_seconds: float = 2.0
     other_handover_max_seconds: float = 1.5
-    max_terminal_overrun_seconds: float = 0.1
+    max_terminal_overrun_seconds: float = 1.1
 
 
 @dataclass(frozen=True)
@@ -61,8 +61,8 @@ def _segment_times(row: dict[str, Any]) -> tuple[float, float]:
         end = _number(row["end_time_sec"], field="end_time_sec")
     else:
         raise ValueError("voice segment lacks clip-local start/end time")
-    if end <= start:
-        raise ValueError(f"voice segment must have positive duration: {start}..{end}")
+    if end < start:
+        raise ValueError(f"voice segment has negative duration: {start}..{end}")
     return start, end
 
 
@@ -111,7 +111,9 @@ def _addressed_to_wearer(
     return any(overlaps)
 
 
-def load_clip_segments(clip: dict[str, Any]) -> tuple[list[SpeechSegment], str]:
+def load_clip_segments(
+    clip: dict[str, Any], *, audit_out: dict[str, Any] | None = None,
+) -> tuple[list[SpeechSegment], str]:
     """Return validated clip-local speaker segments and the wearer person ID."""
     clip_uid = str(clip.get("clip_uid", "")).strip()
     if not clip_uid:
@@ -142,6 +144,7 @@ def load_clip_segments(clip: dict[str, Any]) -> tuple[list[SpeechSegment], str]:
 
     segments: list[SpeechSegment] = []
     seen: set[tuple[str, float, float]] = set()
+    zero_duration: list[dict[str, Any]] = []
     ordinal = 0
     for person in people:
         person_id = _person_key(person.get("person_id"))
@@ -158,6 +161,15 @@ def load_clip_segments(clip: dict[str, Any]) -> tuple[list[SpeechSegment], str]:
             if key in seen:
                 continue
             seen.add(key)
+            if end == start:
+                zero_duration.append({
+                    "person_id": person_id,
+                    "participant_role": "wearer" if person_id == wearer_id else "other",
+                    "time_seconds": start,
+                    "source": dict(source),
+                    "policy": "ignored-zero-duration-artifact-not-expanded",
+                })
+                continue
             segments.append(
                 SpeechSegment(
                     uid=f"{clip_uid}/speech/{ordinal:05d}",
@@ -178,6 +190,12 @@ def load_clip_segments(clip: dict[str, Any]) -> tuple[list[SpeechSegment], str]:
     for left, right in zip(segments, segments[1:]):
         if left.person_id == right.person_id and left.start == right.start and left.end == right.end:
             raise ValueError(f"{clip_uid}: duplicate voice segment after normalization")
+    if audit_out is not None:
+        audit_out.update({
+            "zero_duration_voice_segment_count": len(zero_duration),
+            "zero_duration_voice_segments": zero_duration,
+            "zero_duration_policy": "ignore-with-audit; never clamp or expand",
+        })
     return segments, wearer_id
 
 
@@ -294,9 +312,11 @@ def _base_candidate(
 
 
 def derive_clip_candidates(
-    clip: dict[str, Any], *, split: str, config: CandidateConfig
+    clip: dict[str, Any], *, split: str, config: CandidateConfig,
+    audit_out: dict[str, Any] | None = None,
 ) -> tuple[list[dict[str, Any]], list[dict[str, Any]]]:
-    segments, wearer_id = load_clip_segments(clip)
+    clip_audit: dict[str, Any] = {}
+    segments, wearer_id = load_clip_segments(clip, audit_out=clip_audit)
     declared_clip_end = _number(
         clip.get("clip_end_sec", max((row.end for row in segments), default=0.0)),
         field="clip_end_sec",
@@ -427,7 +447,20 @@ def derive_clip_candidates(
             "terminal_voice_overrun_seconds": terminal_overrun,
             "max_allowed_terminal_overrun_seconds": config.max_terminal_overrun_seconds,
             "terminal_voice_endpoint_preserved_not_clamped": True,
+            "zero_duration_voice_segment_count": clip_audit[
+                "zero_duration_voice_segment_count"
+            ],
+            "zero_duration_voice_segments": clip_audit["zero_duration_voice_segments"],
+            "zero_duration_voice_policy": clip_audit["zero_duration_policy"],
         })
+    clip_audit.update({
+        "declared_clip_end_seconds": declared_clip_end,
+        "maximum_voice_segment_end_seconds": maximum_segment_end,
+        "terminal_voice_overrun_seconds": terminal_overrun,
+        "max_allowed_terminal_overrun_seconds": config.max_terminal_overrun_seconds,
+    })
+    if audit_out is not None:
+        audit_out.update(clip_audit)
     return candidates, [_timeline_row(row) for row in segments]
 
 
@@ -482,7 +515,7 @@ def parse_args() -> argparse.Namespace:
     parser.add_argument("--same-speaker-pause-max-seconds", type=float, default=2.0)
     parser.add_argument("--other-handover-max-seconds", type=float, default=1.5)
     parser.add_argument(
-        "--max-terminal-overrun-seconds", type=float, default=0.1,
+        "--max-terminal-overrun-seconds", type=float, default=1.1,
         help="Allow and audit a bounded official voice endpoint past clip_end_sec (default: %(default)s).",
     )
     parser.add_argument("--overwrite", action="store_true")
@@ -516,10 +549,16 @@ def main() -> None:
     config = CandidateConfig(*numeric)
     rows: list[dict[str, Any]] = []
     timelines: dict[str, list[dict[str, Any]]] = {}
+    clip_audits: dict[str, dict[str, Any]] = {}
     for clip, split in selected:
-        derived, timeline = derive_clip_candidates(clip, split=split, config=config)
+        audit: dict[str, Any] = {}
+        derived, timeline = derive_clip_candidates(
+            clip, split=split, config=config, audit_out=audit
+        )
         rows.extend(derived)
-        timelines[str(clip["clip_uid"])] = timeline
+        clip_uid = str(clip["clip_uid"])
+        timelines[clip_uid] = timeline
+        clip_audits[clip_uid] = audit
     if not rows:
         raise ValueError("selected annotations produced no candidates")
     missing_voice_count_by_clip = {
@@ -545,11 +584,20 @@ def main() -> None:
     _write_jsonl(output, rows)
     _write_json(
         timeline_output,
-        {"schema_version": 1, "dataset": "ego4d_av", "sessions": timelines},
+        {
+            "schema_version": 1,
+            "dataset": "ego4d_av",
+            "sessions": timelines,
+            "clip_audits": clip_audits,
+        },
     )
     counts = Counter(row["candidate_source"] for row in rows)
     print(f"clips={len(selected)} candidates={len(rows)} output={output}")
     print("candidate_sources=" + json.dumps(dict(sorted(counts.items())), ensure_ascii=False))
+    print(
+        "zero_duration_voice_segments_ignored_with_audit="
+        + str(sum(audit["zero_duration_voice_segment_count"] for audit in clip_audits.values()))
+    )
     print(f"timelines={timeline_output}")
 
 

test_prepare_ego4d_av_eot_int_review.py

@@ -129,10 +129,24 @@ class Ego4DAVCandidateTests(unittest.TestCase):
     def test_large_terminal_endpoint_overrun_is_rejected(self) -> None:
         clip = fixture_clip()
         clip["clip_end_sec"] = 11.95
-        clip["camera_wearer"]["voice_segments"].append(segment("wearer", 11.8, 12.2))
+        clip["camera_wearer"]["voice_segments"].append(segment("wearer", 11.8, 13.2))
         with self.assertRaisesRegex(ValueError, "exceeds the audited limit"):
             derive_clip_candidates(clip, split="val", config=CandidateConfig())
 
+    def test_zero_duration_voice_segment_is_ignored_with_explicit_audit(self) -> None:
+        clip = fixture_clip()
+        clip["persons"][1]["voice_segments"].append(segment("p2", 9.0, 9.0))
+        audit: dict[str, object] = {}
+        rows, timeline = derive_clip_candidates(
+            clip, split="val", config=CandidateConfig(), audit_out=audit
+        )
+        self.assertEqual(audit["zero_duration_voice_segment_count"], 1)
+        self.assertFalse(any(row["start_seconds"] == 9.0 for row in timeline))
+        self.assertTrue(all(
+            row["temporal_evidence"]["zero_duration_voice_segment_count"] == 1
+            for row in rows
+        ))
+
 
 if __name__ == "__main__":
     unittest.main()