icassp2027-experiment/docs/easycom_event_review_tool.md

EasyCom wearer-entry/INT pre-label and review pipeline

The workflow has three explicit stages. Speech activity timestamps are read-only throughout; reviewed events never modify the dense 5 Hz activity ground truth.

EasyCom participant speech intervals
  -> floor-episode grouping + rule candidates + structured timing evidence
  -> DeepSeek semantic pre-label (optional but recommended)
  -> human Accept / Relabel / Exclude

1. Generate candidates

Develop the rules on training-side Sessions 4--12:

python3 prepare_easycom_eot_int_review.py \
  --dataset-root /path/to/easycom \
  --sessions Session_4 Session_5 Session_6 Session_7 Session_8 Session_9 Session_10 Session_11 Session_12 \
  --output-dir reports/easycom_floor_event_candidates

The generator emits two independent files:

• easycom_eot_int_candidates.jsonl (legacy-compatible filename) contains every floor-level wearer 0->1 behavior anchor (WEARER_ONSET, WEARER_ENTRY_CANDIDATE), only episode-ending wearer 1->0 boundaries where a new non-wearer onset satisfies wearer_floor_start < other_onset < wearer_offset and the same activity crosses the offset (WEARER_OFFSET, YIELD_CANDIDATE), and hard-negative regions that need no wearer transition: SAME_SPEAKER_PAUSE, OTHER_TO_OTHER, OTHER_BACKCHANNEL, and NON_TAKEOVER_OVERLAP;
• easycom_eot_int_controls.jsonl contains short within-turn raw onsets (WEARER_RESUME) and wearer offsets that fail the strict INT ordering/crossing prerequisite (WEARER_OFFSET_CONTROL). They are retained for audits, but are not sent to DeepSeek or placed in the primary human-review queue.

A raw wearer onset is merged into its preceding floor episode only when its preceding silence is at most 0.8 seconds and no plausible sustained non-wearer takeover occurs in that gap. Short feedback-like lexical or nonverbal activity is recorded but does not by itself split the episode. A non-feedback other contribution that occupies at least continuation_min_seconds of the gap or continues across the resumed onset conservatively preserves a new wearer-entry candidate. Every decision is stored in temporal_evidence, including the raw gap, feedback IDs, competing IDs, threshold, and floor_episode_merged flag.

Entry construction retains every floor-level wearer onset, then searches backward for preceding non-wearer releases as optional context. A release is an admissible option only when the wearer stays silent and no new non-wearer onset or already-active other floor invalidates the interval before the wearer onset. Silence length alone does not invalidate an option. Each wearer onset produces exactly one primary row. A missing release does not demote an entry to a control: an entry may occur from silence or while another speaker is active. entry_context_timing records AFTER_OTHER_RELEASE, OTHER_ACTIVE, or FROM_SILENCE_OR_UNRESOLVED as metadata. When present, opportunity_options contains the only release anchors that DeepSeek or a human may optionally select; neither may invent a timestamp.

Floor-level wearer onsets and strictly qualified offsets carry the rule-stage value candidate; timing alone is not a semantic positive label. Deterministic rows outside the primary queue carry control, never a semantic verdict. The INT prerequisite is wearer_floor_start < other.start < wearer_offset < other.end, with a configurable one-frame (50 ms) tolerance only at the offset boundary. Pre-existing other activity and raw offsets inside a merged wearer episode stay in the controls file. Hard-negative scans are rule hints and still require AI and human review. The JSONL retains candidate source, original event time, participant-level activity, raw durations/gaps/overlap evidence, transcript, and Target_of_Speech when present. Target_of_Speech is annotation assistance only and cannot be a deployable model input.

All heuristic durations are configurable rather than frozen paper rules:

--relation-window-seconds
--same-speaker-pause-max-seconds
--other-handover-max-seconds
--short-vocalization-max-seconds
--continuation-min-seconds
--int-offset-tolerance-seconds
--wearer-resume-merge-max-seconds

Use --max-candidates 1 for a one-event smoke test. summary.json records the exact configuration and candidate counts. easycom_session_timelines.json stores the read-only participant speech lanes used by the full-session web timeline; it is not sent to DeepSeek.

2. Add DeepSeek pre-labels

Prompt v11 first reconstructs the exact participant-resolved chronology and then uses transcript content and Target_of_Speech to determine conversational function. It treats the EasyCom transcript marker [L] as laughter-only evidence, not as a lexical floor take. Wearer laughter during another participant's retained turn is directed toward ENTRY_WEARER_BACKCHANNEL; non-wearer laughter overlapping a retained wearer turn is directed toward INT_OTHER_BACKCHANNEL. Ambiguous laughter falls back to EXCLUDE, and [L] alone is neither evidence of a floor take nor evidence of a visible smile. Preceding release options are optional entry metadata, whereas strict INT ordering remains a deterministic retrieval prerequisite. A positive WEARER_ENTRY may optionally select one supplied release; entries from silence or during other activity use null.

needs_human_review is not requested from DeepSeek: every pre-label enters the human queue by protocol, so the adjudicator assigns that boolean locally. This prevents a string-valued model field from invalidating and retrying an otherwise valid batch.

First inspect exactly what would be uploaded:

python3 adjudicate_easycom_eot_int_candidates.py \
  --input reports/easycom_floor_event_candidates/easycom_eot_int_candidates.jsonl \
  --output reports/easycom_floor_event_candidates/easycom_eot_int_deepseek.jsonl \
  --model deepseek-v4-flash \
  --max-events 1 \
  --dry-run

After setting DEEPSEEK_API_KEY and obtaining authorization to upload the candidate excerpts:

python3 adjudicate_easycom_eot_int_candidates.py \
  --input reports/easycom_floor_event_candidates/easycom_eot_int_candidates.jsonl \
  --output reports/easycom_floor_event_candidates/easycom_eot_int_deepseek.jsonl \
  --model deepseek-v4-flash \
  --allow-external-transcript-upload

DeepSeek sees candidate source, structured timing, per-speaker activity, and transcript together. It returns one of POSITIVE, HARD_NEGATIVE, or EXCLUDE; positives additionally receive WEARER_ENTRY or INT_YIELD, and hard negatives receive a named subtype. It cannot modify timestamps. The output preserves the original rule candidate, the selected preceding-release UID when applicable, and the complete AI decision/provenance. EXCLUDE is the sole non-scoring outcome. It covers out-of-scope non-events (including a normal wearer completion that merely coincides with feedback), insufficient evidence, unresolvable data quality, or disagreement between the two model passes.

Hard negatives use behavior-specific subtypes. ENTRY distinguishes a same-other mid-turn pause, a different-other transition, and a wearer backchannel that does not take the floor. INT first distinguishes an other backchannel, then a non-backchannel overlap after which the wearer demonstrably holds or resumes the floor, and finally another interpretable overlap that fails to become a sustained takeover. OTHER_HARD_NEGATIVE is reserved for valid false-positive opportunities outside these named behaviors; it is never an uncertainty fallback. A short but complete answer to an addressed question remains a positive wearer entry; duration alone and later resumption by another speaker do not turn it into a wearer backchannel.

During live adjudication, the CLI displays a dependency-free progress bar over all candidate/pass work, including the current pass, batch, percentage, and elapsed time. Redirected logs emit periodic progress lines instead of carriage return animation.

As defense in depth, the adjudicator also detects and skips non-crossing WEARER_OFFSET rows from older v2 candidate files before making any API call.

3. Human review

Run the web workspace on the AI-enriched JSONL:

python3 review_floor_events.py \
  --dataset-root /path/to/easycom \
  --events reports/easycom_floor_event_candidates/easycom_eot_int_deepseek.jsonl \
  --output reports/easycom_event_reviews/reviews.jsonl

Open http://127.0.0.1:8765. The page shows a session candidate overview, participant activity near the event, video/audio, transcript, structured timing, and the AI verdict/confidence/reason. A decision guide first states the behavioral question, then presents the exact AI verdict/subtype with a plain-language meaning; the Accept control repeats that exact label so its effect is explicit. The common path is one operation:

• Accept keeps the AI verdict/type and optional preceding-release metadata;
• Relabel expands verdict and subtype controls; a positive entry may optionally select one of the supplied release anchors;
• Exclude requires a concrete out-of-scope or data-quality reason.

The current primary-review UI deliberately has no second-review action. Older saved second_review records remain readable for backward compatibility.

Boundary correction is hidden under an optional disclosure and is limited to one 200 ms bin. The original timestamp is always retained. Each save atomically rewrites JSONL and CSV while preserving the rule candidate, AI decision, human decision, relabel flag, exclusion state, confidence, and provenance. The current UI does not ask for a reviewer name; the export retains an empty reviewer field only for compatibility with older review files.

The primary review queue also filters legacy non-crossing WEARER_OFFSET rows and explicit WEARER_RESUME controls. To audit the controls intentionally, pass --events .../easycom_eot_int_controls.jsonl --include-controls and use a separate review output file. The old --include-offset-controls spelling remains an alias.

Protocol wording for the paper appendix:

Raw wearer speech onsets are first grouped into floor-level speech episodes by merging short within-turn resumptions that are not separated by a competing floor takeover. Every resulting floor-level 0→1 transition is proposed as a wearer-entry candidate; a preceding non-wearer release is retained as context rather than required for positive status.

The original speech intervals are preserved unchanged for dense forecasting and auditability; this grouping is used only for behavioral event annotation.

Freeze discipline

Develop the guide and resolve disagreements only on Sessions 4--12. Freeze categories and thresholds before reviewing official test Sessions 1--3. Use two independent human review files plus a separate adjudication output for paper- quality labels. DeepSeek is pre-label metadata, not an independent reviewer.