code/scenarios.json

[
 {"name": "01_igbo_greetings_english", "desc": "Mostly English with Igbo greetings/closing (ACE's target case). Includes a long turn (backchannel) and a barge-in. User turns are TTS-synthesized (Flash, female Igbo voice).",
  "turns": [
   {"label": "TTS-synthesized user turn", "voice": "waxal_ibo_5_female", "text": "<ibo>Ndewo!</ibo> <eng>Good morning. Please, can you help me plan my day? I have a lot to do.</eng>"},
   {"label": "TTS-synthesized user turn (long, mid-turn pause)", "voice": "waxal_ibo_5_female", "text_parts": ["<eng>I need to go to the bank, and then the market,</eng>", "<eng>and in the evening I also want to call my mother.</eng>"], "inner_pause": 0.4},
   {"label": "TTS-synthesized user turn (barge-in 1.5 s into the reply)", "voice": "waxal_ibo_5_female", "text": "<eng>Wait, sorry, the bank closes early today.</eng>", "barge_at": 1.5},
   {"label": "TTS-synthesized user turn", "voice": "waxal_ibo_5_female", "text": "<eng>Okay, thank you so much.</eng> <ibo>Daalụ!</ibo>"}]},
 {"name": "02_yoruba_greetings_english", "desc": "Yoruba greetings + English. User turns TTS-synthesized (Flash, female Yoruba voice).",
  "turns": [
   {"label": "TTS-synthesized user turn", "voice": "waxal_yor_5_female", "text": "<yor>Ẹ káàárọ̀!</yor> <eng>I want to buy a new phone for my sister.</eng>"},
   {"label": "TTS-synthesized user turn", "voice": "waxal_yor_5_female", "text": "<eng>Her budget is about one hundred thousand naira.</eng>"},
   {"label": "TTS-synthesized user turn", "voice": "waxal_yor_5_female", "text": "<eng>Okay, that sounds good.</eng> <yor>Ẹ ṣé!</yor>"}]},
 {"name": "03_pidgin_bargein", "desc": "Nigerian Pidgin throughout, with a barge-in. User turns TTS-synthesized (Flash, female voice).",
  "turns": [
   {"label": "TTS-synthesized user turn", "voice": "waxal_yor_5_female", "text": "<pcm>How far? Abeg, wetin I fit cook tonight wey no go take long?</pcm>"},
   {"label": "TTS-synthesized user turn (barge-in 1.2 s into the reply)", "voice": "waxal_yor_5_female", "text": "<pcm>No be rice o, I no get rice for house.</pcm>", "barge_at": 1.2},
   {"label": "TTS-synthesized user turn", "voice": "waxal_yor_5_female", "text": "<pcm>Okay, na so. Thank you well well.</pcm>"}]},
 {"name": "04_hausa_greetings_english", "desc": "Hausa greeting + English, Hausa closing. User turns TTS-synthesized (Flash, female voice).",
  "turns": [
   {"label": "TTS-synthesized user turn", "voice": "waxal_ibo_5_female", "text": "<hau>Sannu, barka da yamma.</hau> <eng>I am travelling to Kano tomorrow.</eng>"},
   {"label": "TTS-synthesized user turn", "voice": "waxal_ibo_5_female", "text": "<eng>What should I pack for the weather there?</eng>"},
   {"label": "TTS-synthesized user turn", "voice": "waxal_ibo_5_female", "text": "<hau>To, na gode.</hau>"}]},
 {"name": "05_igbo_heavy", "desc": "Igbo-heavier user (shows the weak spot: LLM Igbo beyond phrases). User turns TTS-synthesized (Flash, female Igbo voice).",
  "turns": [
   {"label": "TTS-synthesized user turn", "voice": "waxal_ibo_5_female", "text": "<ibo>Kedu? Biko, nyere m aka.</ibo> <eng>I need a taxi to the airport.</eng>"},
   {"label": "TTS-synthesized user turn", "voice": "waxal_ibo_5_female", "text": "<ibo>Ee, ọ dị mma.</ibo> <eng>I will leave at six.</eng>"},
   {"label": "TTS-synthesized user turn", "voice": "waxal_ibo_5_female", "text": "<ibo>Daalụ, ka ọ dị.</ibo>"}]},
 {"name": "06_real_waxal_clips", "desc": "User turns are REAL held-out WAXAL test recordings (CC-BY-SA-4.0; read sentences, so the content is not conversational): Pidgin, an English+Igbo concatenation, Hausa.",
  "turns": [
   {"label": "real WAXAL test clip pcm_121", "wav": "/scratch/duplex/asr_eval/wav/pcm_121.wav"},
   {"label": "real WAXAL/FLEURS clips concatenated (eng + ibo), cs_000", "wav": "/scratch/duplex/asr_eval/wav/cs_000.wav"},
   {"label": "real WAXAL test clip hau_081", "wav": "/scratch/duplex/asr_eval/wav/hau_081.wav"}]}
]