9. 화자별 발화 구조 및 연관성 상세보고서
논문 작성용 상세 본문
현재 9번 분석은 화자 후보별 발화량, 발화 비중, 턴 수, 평균 세그먼트 길이, 응답쌍과 같은 구조적 지표를 다룬다.
S0 발화 이후 S1 반응을 시간 순서 기반 연관성 후보로 분석할 수 있으나, segment-level stress_score가 없어 stress 전이 또는 고스트레스 반응은 아직 계산할 수 없다.
따라서 운영기와 문서에는 READY 구조 지표와 PENDING stress 지표를 동시에 표시한다.
현재 화자별 분석은 발화시간, 발화 비중, 세그먼트 수, 턴 수, 평균 구간 길이, 응답쌍과 같은 구조 지표를 중심으로 수행한다.
이 지표는 통화 내부 상호작용의 형태를 설명하지만 감정 또는 스트레스의 방향을 직접 의미하지 않는다.
segment-level stress_score가 0건이므로 S0 stress와 S1 stress의 평균, 전이, 고스트레스 반응은 모두 PENDING으로 유지한다.
분석 수치 및 결과표
표 9-1. 분석 가능 지표
| 지표 | 상태 | 해석 |
|---|---|---|
| 화자별 발화시간 | READY | 구조 비교 |
| 화자별 발화 비중 | READY | 구조 비교 |
| 턴 수/세그먼트 수 | READY | 상호작용 강도 후보 |
| S0→S1 응답쌍 | READY | 시간 순서 후보 |
표 9-2. 미완료 stress 지표
| 지표 | 상태 | 필요 조건 |
|---|---|---|
| S0 stress 평균 | PENDING | segment STT/WPM/stress |
| S1 stress 평균 | PENDING | segment STT/WPM/stress |
| S0 stress→S1 stress | PENDING | 응답쌍 stress 병합 |
| high-stress segment | PENDING | stress_score 완료 |
표 9-Q. 연구 질문
| 번호 | 연구 질문 |
|---|---|
| 1 | 화자 후보별 발화량·비중·턴·응답 구조는 어떤 패턴을 보이는가? |
| 2 | stress 관련 분석은 현재 어디까지 가능한가? |
표 9-M. 분석 방법 및 도구
| 순서 | 방법/도구 |
|---|---|
| 1 | Speaker descriptive statistics |
| 2 | Turn and speech-ratio analysis |
| 3 | Sequential association candidates |
| 4 | READY/PENDING separation |
분석 그림 및 도식
원본 분석 산출물 연계
산출물 CSV: research_continuity/26_speaker_level_stress_interaction_v201/speaker_interaction_types_v201.csv (행 1,004, 열 4)
대용량 표 요약·앞/뒤 표본 포함, 원본은 데이터 ZIP 수록
| call_id | interaction_type | basis | role_mapping |
|---|---|---|---|
| __SUMMARY__ | Type A: balanced speech ratio | count | 179 |
| __SUMMARY__ | Type B: SPEAKER_00 dominant speech ratio | count | 434 |
| __SUMMARY__ | Type C: SPEAKER_01 dominant speech ratio | count | 347 |
| __SUMMARY__ | Type E: SPEAKER_01 zero-slot | count | 40 |
| 00__350002022300000.wav | Type A: balanced speech ratio | speech_ratio_gap | pending |
| 00__350002022300001.wav | Type A: balanced speech ratio | speech_ratio_gap | pending |
| 00__350002022300002.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 00__350002022300003.wav | Type A: balanced speech ratio | speech_ratio_gap | pending |
| 00__350002022300004.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 00__350002022300005.wav | Type C: SPEAKER_01 dominant speech ratio | speech_ratio_gap | pending |
| 00__350002022300006.wav | Type A: balanced speech ratio | speech_ratio_gap | pending |
| 00__350002022300007.wav | Type A: balanced speech ratio | speech_ratio_gap | pending |
| 00__350002022300008.wav | Type A: balanced speech ratio | speech_ratio_gap | pending |
| 00__350002022300009.wav | Type A: balanced speech ratio | speech_ratio_gap | pending |
| 00__350002022300010.wav | Type E: SPEAKER_01 zero-slot | speech_ratio_gap | pending |
| 00__350002022300011.wav | Type A: balanced speech ratio | speech_ratio_gap | pending |
| 00__350002022300012.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 00__350002022300013.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 00__350002022300014.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 00__350002022300015.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 00__350002022300016.wav | Type C: SPEAKER_01 dominant speech ratio | speech_ratio_gap | pending |
| 00__350002022300017.wav | Type C: SPEAKER_01 dominant speech ratio | speech_ratio_gap | pending |
| 00__350002022300018.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 00__350002022300019.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 00__350002022300020.wav | Type C: SPEAKER_01 dominant speech ratio | speech_ratio_gap | pending |
| ... 중간 954행은 전체 데이터 부록 ZIP에 원본으로 포함 ... | |||
| 09__350002022300975.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300976.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300977.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300978.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300979.wav | Type E: SPEAKER_01 zero-slot | speech_ratio_gap | pending |
| 09__350002022300980.wav | Type C: SPEAKER_01 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300981.wav | Type A: balanced speech ratio | speech_ratio_gap | pending |
| 09__350002022300982.wav | Type E: SPEAKER_01 zero-slot | speech_ratio_gap | pending |
| 09__350002022300983.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300984.wav | Type E: SPEAKER_01 zero-slot | speech_ratio_gap | pending |
| 09__350002022300985.wav | Type C: SPEAKER_01 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300986.wav | Type A: balanced speech ratio | speech_ratio_gap | pending |
| 09__350002022300987.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300988.wav | Type C: SPEAKER_01 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300989.wav | Type A: balanced speech ratio | speech_ratio_gap | pending |
| 09__350002022300990.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300991.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300992.wav | Type A: balanced speech ratio | speech_ratio_gap | pending |
| 09__350002022300993.wav | Type A: balanced speech ratio | speech_ratio_gap | pending |
| 09__350002022300994.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300995.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300996.wav | Type B: SPEAKER_00 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300997.wav | Type C: SPEAKER_01 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300998.wav | Type C: SPEAKER_01 dominant speech ratio | speech_ratio_gap | pending |
| 09__350002022300999.wav | Type C: SPEAKER_01 dominant speech ratio | speech_ratio_gap | pending |
산출물 CSV: research_continuity/26_speaker_level_stress_interaction_v201/speaker_sequential_association_v201.csv (행 8, 열 6)
전체 행 포함
| pair_metric | correlation | p_value | n | interpretation | claim_level |
|---|---|---|---|---|---|
| S0 total speech sec → S1 total speech sec | 0.341355 | 1.26501e-27 | 960 | call-level speech amount association | SEQUENTIAL_ASSOCIATION_NOT_CAUSAL |
| S0 turn count → S1 turn count | 0.862617 | 8.5856e-286 | 960 | call-level turn count association | SEQUENTIAL_ASSOCIATION_NOT_CAUSAL |
| S0 mean turn duration → S1 mean turn duration | -0.372985 | 4.72794e-33 | 960 | call-level mean turn duration association | SEQUENTIAL_ASSOCIATION_NOT_CAUSAL |
| S0 segment duration → next S1 segment duration | -0.082305 | 1.64032e-43 | 28149 | response-pair sequential association | SEQUENTIAL_ASSOCIATION_NOT_CAUSAL |
| S0 segment duration → S1 response latency | 0.078083 | 2.53106e-39 | 28149 | response latency association | SEQUENTIAL_ASSOCIATION_NOT_CAUSAL |
| S0 stress → S1 stress | 0 | PENDING: segment-level acoustic/STT/stress features are required. Current v201 keeps this as a safe analysis slot, not a claim. | PENDING_FEATURE_MERGE | ||
| S0 energy → S1 response latency | 0 | PENDING: segment-level acoustic/STT/stress features are required. Current v201 keeps this as a safe analysis slot, not a claim. | PENDING_FEATURE_MERGE | ||
| S0 wpm → S1 wpm | 0 | PENDING: segment-level acoustic/STT/stress features are required. Current v201 keeps this as a safe analysis slot, not a claim. | PENDING_FEATURE_MERGE |
결과 해석
현재 결과는 상호작용 구조의 기술통계와 시간 순서 후보 분석으로 활용할 수 있다.
한계 및 논문 반영 기준
stress 관련 결론과 고객/상담사 역할 기반 결론은 보류한다.