| """ |
| Add French, Spanish and German entries to the AMD dataset. |
| |
| Adds ~150 samples per language (600 total) across all 4 classes: |
| - human: ~50 real (MINDS14) + ~25 TTS per language |
| - voicemail/ivr/am: ~25 TTS each per language |
| |
| French gets extra weight (200 samples) since it's the primary EU market. |
| """ |
|
|
| import asyncio |
| import edge_tts |
| import numpy as np |
| import soundfile as sf |
| import os |
| import sys |
| import random |
| import subprocess |
| from pydub import AudioSegment |
| from scipy.signal import butter, lfilter |
| from datasets import load_dataset, Audio, Dataset, DatasetDict, ClassLabel, concatenate_datasets |
| from huggingface_hub import hf_hub_download |
| import pyarrow.parquet as pq |
|
|
| random.seed(123) |
| np.random.seed(123) |
| sys.stdout = os.fdopen(sys.stdout.fileno(), 'w', buffering=1) |
|
|
| SAMPLE_RATE = 16000 |
| MAX_LENGTH_S = 10.0 |
| OUTPUT_DIR = "/app/amd_multilingual" |
| audio_dir = os.path.join(OUTPUT_DIR, "audio") |
| os.makedirs(audio_dir, exist_ok=True) |
|
|
| LABELS = ["human", "voicemail", "ivr", "answering_machine"] |
|
|
| |
| |
| |
| FR_NAMES = ["Jean", "Marie", "Pierre", "Sophie", "Nicolas", "Isabelle", "François", "Claire", |
| "Laurent", "Nathalie", "Thierry", "Catherine", "Philippe", "Valérie"] |
| FR_COMPANIES = ["France Télécom", "Société Générale", "Crédit Agricole", "AXA Assurance", |
| "Bouygues", "Orange", "Carrefour", "SNCF"] |
|
|
| FR_HUMAN = [ |
| "Allô ?", "Oui, bonjour.", "Allô, qui est à l'appareil ?", |
| "Bonjour, {name} à l'appareil.", "Oui, c'est {name}.", |
| "Allô ? Oui, je vous écoute.", "Bonjour, comment puis-je vous aider ?", |
| "Oui bonjour, {name} de {company}.", "Allô ? Attendez une seconde s'il vous plaît.", |
| "Oui, je suis là. Allez-y.", "Bonjour, qu'est-ce que je peux faire pour vous ?", |
| "Allô ? Oui, c'est bien {name}.", "Bonjour, je vous attendais.", |
| "Oui, excusez-moi, j'étais en réunion.", "Allô ? Désolé, mauvaise connexion.", |
| "Oui bonjour, c'est à quel sujet ?", "Bien sûr, je vous écoute.", |
| "Bonjour, merci de rappeler.", "Oui, un instant je vérifie.", |
| ] |
|
|
| FR_VOICEMAIL = [ |
| "Bonjour, vous êtes bien sur le répondeur de {name}. Je ne suis pas disponible pour le moment. Laissez un message après le bip.", |
| "Salut, c'est {name}. Je ne peux pas répondre pour l'instant. Laissez-moi un message.", |
| "Vous êtes sur la messagerie de {name} chez {company}. Merci de laisser vos coordonnées.", |
| "Bonjour, {name} à l'appareil. Je suis absent du bureau. Veuillez laisser un message.", |
| "Bonjour, vous avez atteint la boîte vocale de {name}. Merci de laisser un message après le signal sonore.", |
| "Salut, c'est {name}. Désolé de ne pas pouvoir vous répondre. Laissez un message et je vous rappelle.", |
| "Bonjour, je suis {name}. Je ne suis pas joignable actuellement. Laissez votre nom et numéro après le bip.", |
| "Vous êtes bien chez {name}. Je suis en déplacement. Merci de laisser un message.", |
| ] |
|
|
| FR_IVR = [ |
| "Merci d'appeler {company}. Pour le service commercial, tapez 1. Pour le support technique, tapez 2. Pour un conseiller, tapez 0.", |
| "Bienvenue chez {company}. Pour le français, appuyez sur 1. For English, press 2.", |
| "Merci de patienter, votre appel est important pour nous. Un conseiller va vous répondre.", |
| "Tapez 1 pour vos factures. Tapez 2 pour le service client. Tapez 3 pour les réclamations.", |
| "Bienvenue sur le serveur vocal de {company}. Si vous connaissez le numéro de poste, composez-le maintenant.", |
| "Votre appel est en attente. Temps d'attente estimé : cinq minutes. Pour laisser un message, tapez 1.", |
| "Merci d'appeler {company}. Nos horaires sont du lundi au vendredi, de 9 heures à 18 heures.", |
| "Cet appel est susceptible d'être enregistré à des fins de qualité. Tapez 1 pour continuer.", |
| ] |
|
|
| FR_AM = [ |
| "Le correspondant que vous cherchez à joindre n'est pas disponible. Veuillez laisser un message après le signal sonore.", |
| "Le numéro que vous avez composé n'est pas attribué. Veuillez vérifier le numéro.", |
| "La boîte vocale est pleine. Veuillez rappeler ultérieurement.", |
| "Votre correspondant est actuellement indisponible. Laissez un message après le bip.", |
| "Vous êtes redirigé vers la messagerie vocale. Laissez votre message après le signal.", |
| "Le correspondant ne peut pas prendre votre appel. Veuillez laisser un message.", |
| ] |
|
|
| |
| |
| |
| ES_NAMES = ["Carlos", "María", "José", "Ana", "Miguel", "Carmen", "Antonio", "Laura", |
| "Pablo", "Isabel", "Diego", "Sofía"] |
| ES_COMPANIES = ["Telefónica", "Banco Santander", "BBVA", "Mapfre Seguros", |
| "Iberdrola", "Repsol"] |
|
|
| ES_HUMAN = [ |
| "¿Aló?", "Sí, buenos días.", "¿Quién habla?", |
| "Hola, soy {name}.", "Buenos días, {name} al habla.", |
| "Sí, dígame.", "Hola, ¿en qué puedo ayudarle?", |
| "Buenos días, {company}, habla {name}.", "Un momento por favor.", |
| "Sí, aquí estoy. Adelante.", "¿Hola? Sí, soy yo.", |
| "Hola, estaba esperando su llamada.", "Disculpe, estaba en una reunión.", |
| "Sí, claro, le escucho.", "Buenos días, gracias por devolver la llamada.", |
| ] |
|
|
| ES_VOICEMAIL = [ |
| "Hola, has llamado a {name}. No estoy disponible en este momento. Deja un mensaje después del tono.", |
| "Hola, soy {name}. No puedo atender tu llamada ahora. Deja un mensaje.", |
| "Has contactado con {name} de {company}. Deja tu nombre y número después de la señal.", |
| "Hola, soy {name}. Estoy fuera de la oficina. Por favor deja un mensaje.", |
| "El buzón de voz de {name}. Deja tu mensaje después del tono.", |
| "Hola, no puedo contestar ahora. Deja un mensaje y te llamo de vuelta.", |
| ] |
|
|
| ES_IVR = [ |
| "Gracias por llamar a {company}. Para ventas, pulse 1. Para soporte técnico, pulse 2. Para un operador, pulse 0.", |
| "Bienvenido a {company}. Para español, pulse 1. For English, press 2.", |
| "Su llamada es importante para nosotros. Por favor espere en línea.", |
| "Pulse 1 para facturación. Pulse 2 para atención al cliente. Pulse 3 para reclamaciones.", |
| "Gracias por llamar a {company}. Nuestro horario es de lunes a viernes, de 9 a 18 horas.", |
| "Esta llamada puede ser grabada con fines de calidad. Pulse 1 para continuar.", |
| ] |
|
|
| ES_AM = [ |
| "La persona a la que llama no está disponible. Deje un mensaje después de la señal.", |
| "El número marcado no existe. Por favor verifique el número.", |
| "El buzón de voz está lleno. Intente más tarde.", |
| "Su llamada ha sido redirigida al buzón de voz. Deje su mensaje después del tono.", |
| "El cliente al que llama no está disponible. Por favor deje un mensaje.", |
| ] |
|
|
| |
| |
| |
| DE_NAMES = ["Thomas", "Anna", "Michael", "Sabine", "Stefan", "Claudia", "Andreas", "Monika", |
| "Peter", "Katharina", "Wolfgang", "Birgit"] |
| DE_COMPANIES = ["Deutsche Telekom", "Commerzbank", "Allianz", "Siemens", |
| "BMW Service", "Deutsche Bahn"] |
|
|
| DE_HUMAN = [ |
| "Hallo?", "Ja, guten Tag.", "Wer spricht bitte?", |
| "Hallo, hier ist {name}.", "Guten Tag, {name} am Apparat.", |
| "Ja bitte?", "Hallo, wie kann ich Ihnen helfen?", |
| "Guten Tag, {company}, hier spricht {name}.", "Einen Moment bitte.", |
| "Ja, ich bin dran. Bitte.", "Hallo? Ja, ich bin es.", |
| "Hallo, ich habe Ihren Anruf erwartet.", "Entschuldigung, ich war in einer Besprechung.", |
| "Ja, natürlich, ich höre.", "Guten Tag, danke für den Rückruf.", |
| ] |
|
|
| DE_VOICEMAIL = [ |
| "Hallo, Sie sind mit dem Anrufbeantworter von {name} verbunden. Bitte hinterlassen Sie eine Nachricht nach dem Signalton.", |
| "Hallo, hier ist {name}. Ich kann gerade nicht ans Telefon gehen. Bitte hinterlassen Sie eine Nachricht.", |
| "Sie haben {name} bei {company} erreicht. Bitte hinterlassen Sie Ihren Namen und Ihre Nummer.", |
| "Hallo, ich bin {name}. Ich bin derzeit nicht erreichbar. Bitte sprechen Sie nach dem Ton.", |
| "Die Mailbox von {name}. Bitte hinterlassen Sie eine Nachricht nach dem Signalton.", |
| "Hallo, leider kann ich Ihren Anruf nicht entgegennehmen. Bitte hinterlassen Sie mir eine Nachricht.", |
| ] |
|
|
| DE_IVR = [ |
| "Vielen Dank für Ihren Anruf bei {company}. Für den Vertrieb drücken Sie die 1. Für den Kundendienst drücken Sie die 2.", |
| "Willkommen bei {company}. Für Deutsch drücken Sie die 1. For English, press 2.", |
| "Ihr Anruf ist uns wichtig. Bitte bleiben Sie in der Leitung.", |
| "Drücken Sie die 1 für Rechnungen. Drücken Sie die 2 für den Kundenservice. Drücken Sie die 3 für Beschwerden.", |
| "Vielen Dank für Ihren Anruf bei {company}. Unsere Öffnungszeiten sind Montag bis Freitag, 9 bis 18 Uhr.", |
| "Dieser Anruf kann zu Qualitätszwecken aufgezeichnet werden. Drücken Sie die 1 um fortzufahren.", |
| ] |
|
|
| DE_AM = [ |
| "Der gewünschte Teilnehmer ist nicht erreichbar. Bitte hinterlassen Sie eine Nachricht nach dem Signalton.", |
| "Die gewählte Rufnummer ist nicht vergeben. Bitte überprüfen Sie die Nummer.", |
| "Die Mailbox ist voll. Bitte versuchen Sie es später erneut.", |
| "Ihr Anruf wird an die Mailbox weitergeleitet. Bitte sprechen Sie nach dem Ton.", |
| "Der Teilnehmer ist vorübergehend nicht erreichbar. Bitte hinterlassen Sie eine Nachricht.", |
| ] |
|
|
| |
| |
| |
| LANG_CONFIG = { |
| "fr": { |
| "human_voices": [ |
| ("fr-FR-DeniseNeural", "+5%", "+5Hz"), |
| ("fr-FR-RemyMultilingualNeural", "+0%", "+0Hz"), |
| ("fr-FR-VivienneMultilingualNeural", "+3%", "+5Hz"), |
| ("fr-CA-SylvieNeural", "+5%", "+5Hz"), |
| ("fr-CA-ThierryNeural", "+0%", "+0Hz"), |
| ("fr-BE-CharlineNeural", "+5%", "+0Hz"), |
| ("fr-BE-GerardNeural", "+0%", "+0Hz"), |
| ], |
| "voicemail_voices": [ |
| ("fr-FR-DeniseNeural", "-5%", "+0Hz"), |
| ("fr-FR-RemyMultilingualNeural", "-5%", "-10Hz"), |
| ("fr-CA-SylvieNeural", "-5%", "+0Hz"), |
| ("fr-BE-GerardNeural", "-5%", "-5Hz"), |
| ], |
| "ivr_voices": [ |
| ("fr-FR-RemyMultilingualNeural", "-20%", "-50Hz"), |
| ("fr-FR-DeniseNeural", "-20%", "-30Hz"), |
| ("fr-CA-AntoineNeural", "-20%", "-40Hz"), |
| ], |
| "am_voices": [ |
| ("fr-FR-RemyMultilingualNeural", "-15%", "-20Hz"), |
| ("fr-FR-DeniseNeural", "-15%", "-25Hz"), |
| ], |
| "names": FR_NAMES, "companies": FR_COMPANIES, |
| "human": FR_HUMAN, "voicemail": FR_VOICEMAIL, "ivr": FR_IVR, "am": FR_AM, |
| "minds14_parquet": "fr-FR/train-00000-of-00001.parquet", |
| "n_real_human": 75, |
| "n_tts_human": 50, |
| "n_voicemail": 40, |
| "n_ivr": 40, |
| "n_am": 40, |
| }, |
| "es": { |
| "human_voices": [ |
| ("es-ES-ElviraNeural", "+5%", "+5Hz"), |
| ("es-ES-AlvaroNeural", "+0%", "+0Hz"), |
| ("es-ES-XimenaNeural", "+3%", "+5Hz"), |
| ("es-MX-DaliaNeural", "+5%", "+5Hz"), |
| ("es-MX-JorgeNeural", "+0%", "+0Hz"), |
| ], |
| "voicemail_voices": [ |
| ("es-ES-ElviraNeural", "-5%", "+0Hz"), |
| ("es-ES-AlvaroNeural", "-5%", "-10Hz"), |
| ("es-MX-DaliaNeural", "-5%", "+0Hz"), |
| ], |
| "ivr_voices": [ |
| ("es-ES-AlvaroNeural", "-20%", "-50Hz"), |
| ("es-ES-ElviraNeural", "-20%", "-30Hz"), |
| ], |
| "am_voices": [ |
| ("es-ES-AlvaroNeural", "-15%", "-20Hz"), |
| ], |
| "names": ES_NAMES, "companies": ES_COMPANIES, |
| "human": ES_HUMAN, "voicemail": ES_VOICEMAIL, "ivr": ES_IVR, "am": ES_AM, |
| "minds14_parquet": "es-ES/train-00000-of-00001.parquet", |
| "n_real_human": 50, |
| "n_tts_human": 25, |
| "n_voicemail": 25, |
| "n_ivr": 25, |
| "n_am": 25, |
| }, |
| "de": { |
| "human_voices": [ |
| ("de-DE-AmalaNeural", "+5%", "+5Hz"), |
| ("de-DE-ConradNeural", "+0%", "+0Hz"), |
| ("de-DE-FlorianMultilingualNeural", "+3%", "+5Hz"), |
| ("de-DE-SeraphinaMultilingualNeural", "+5%", "+0Hz"), |
| ("de-AT-IngridNeural", "+0%", "+0Hz"), |
| ("de-AT-JonasNeural", "+0%", "+0Hz"), |
| ], |
| "voicemail_voices": [ |
| ("de-DE-AmalaNeural", "-5%", "+0Hz"), |
| ("de-DE-ConradNeural", "-5%", "-10Hz"), |
| ("de-AT-IngridNeural", "-5%", "+0Hz"), |
| ], |
| "ivr_voices": [ |
| ("de-DE-ConradNeural", "-20%", "-50Hz"), |
| ("de-DE-FlorianMultilingualNeural", "-20%", "-40Hz"), |
| ], |
| "am_voices": [ |
| ("de-DE-ConradNeural", "-15%", "-20Hz"), |
| ], |
| "names": DE_NAMES, "companies": DE_COMPANIES, |
| "human": DE_HUMAN, "voicemail": DE_VOICEMAIL, "ivr": DE_IVR, "am": DE_AM, |
| "minds14_parquet": "de-DE/train-00000-of-00001.parquet", |
| "n_real_human": 50, |
| "n_tts_human": 25, |
| "n_voicemail": 25, |
| "n_ivr": 25, |
| "n_am": 25, |
| }, |
| } |
|
|
| |
| |
| |
| def telephone_bandpass(audio, sr=16000): |
| nyq = sr / 2 |
| b, a = butter(4, [300 / nyq, min(3400 / nyq, 0.99)], btype='band') |
| return lfilter(b, a, audio).astype(np.float32) |
|
|
| def add_noise(audio, snr_db=25): |
| p = np.mean(audio ** 2) + 1e-10 |
| return (audio + np.random.normal(0, np.sqrt(p / (10 ** (snr_db / 10))), len(audio))).astype(np.float32) |
|
|
| def generate_tone(freq, dur, sr=16000, amp=0.3): |
| t = np.arange(int(dur * sr)) / sr |
| return (amp * np.sin(2 * np.pi * freq * t)).astype(np.float32) |
|
|
| def generate_dtmf(digit, dur=0.15, sr=16000, amp=0.15): |
| dtmf = {'1':(697,1209),'2':(697,1336),'3':(697,1477),'4':(770,1209), |
| '5':(770,1336),'6':(770,1477),'7':(852,1209),'8':(852,1336), |
| '9':(852,1477),'0':(941,1336)} |
| f1, f2 = dtmf.get(digit, (697, 1209)) |
| t = np.arange(int(dur * sr)) / sr |
| return (amp * (np.sin(2*np.pi*f1*t) + np.sin(2*np.pi*f2*t))).astype(np.float32) |
|
|
| def silence(dur, sr=16000): |
| return np.random.normal(0, 0.001, int(dur * sr)).astype(np.float32) |
|
|
| def telephony_fx(audio, sr=16000): |
| audio = telephone_bandpass(audio, sr) |
| audio = add_noise(audio, snr_db=np.random.uniform(18, 35)) |
| return np.clip(audio * np.random.uniform(0.6, 1.3), -1, 1).astype(np.float32) |
|
|
| def load_mp3_wav(path, sr=16000): |
| a = AudioSegment.from_mp3(path).set_frame_rate(sr).set_channels(1).set_sample_width(2) |
| return np.array(a.get_array_of_samples(), dtype=np.float32) / 32768.0 |
|
|
| def decode_audio_bytes(audio_bytes, target_sr=16000): |
| proc = subprocess.run( |
| ['ffmpeg', '-i', 'pipe:0', '-f', 's16le', '-ar', str(target_sr), '-ac', '1', 'pipe:1'], |
| input=audio_bytes, capture_output=True) |
| if proc.returncode == 0 and len(proc.stdout) > 0: |
| return np.frombuffer(proc.stdout, dtype=np.int16).astype(np.float32) / 32768.0 |
| return None |
|
|
| def fill(template, names, companies): |
| return template.format(name=random.choice(names), company=random.choice(companies)) |
|
|
| async def gen_tts(text, voice, rate, pitch, wav_path): |
| mp3 = wav_path.replace('.wav', '.mp3') |
| await edge_tts.Communicate(text, voice=voice, rate=rate, pitch=pitch).save(mp3) |
| audio = load_mp3_wav(mp3, SAMPLE_RATE) |
| max_s = int(MAX_LENGTH_S * SAMPLE_RATE) |
| audio = telephony_fx(audio[:max_s] if len(audio) > max_s else audio, SAMPLE_RATE) |
| sf.write(wav_path, audio, SAMPLE_RATE) |
| os.remove(mp3) |
| return audio |
|
|
| |
| |
| |
| async def main(): |
| all_paths = [] |
| all_labels = [] |
| file_idx = 0 |
|
|
| for lang_code, cfg in LANG_CONFIG.items(): |
| print(f"\n{'='*60}") |
| print(f" LANGUAGE: {lang_code.upper()}") |
| print(f"{'='*60}") |
|
|
| names = cfg["names"] |
| companies = cfg["companies"] |
|
|
| |
| n_real = cfg["n_real_human"] |
| print(f"\n Loading {n_real} real human samples from MINDS14...") |
| try: |
| local = hf_hub_download('PolyAI/minds14', cfg["minds14_parquet"], repo_type='dataset') |
| table = pq.read_table(local) |
| audio_col = table.column('audio') |
| max_s = int(MAX_LENGTH_S * SAMPLE_RATE) |
| loaded = 0 |
| for i in range(len(audio_col)): |
| if loaded >= n_real: |
| break |
| row = audio_col[i].as_py() |
| audio_bytes = row.get('bytes') |
| if not audio_bytes: |
| continue |
| arr = decode_audio_bytes(audio_bytes, SAMPLE_RATE) |
| if arr is None or len(arr) < SAMPLE_RATE: |
| continue |
| if len(arr) > max_s: |
| start = random.randint(0, len(arr) - max_s) |
| arr = arr[start:start + max_s] |
| arr = add_noise(arr, snr_db=np.random.uniform(22, 35)) |
| p = os.path.join(audio_dir, f"{lang_code}_human_real_{file_idx:05d}.wav") |
| sf.write(p, arr, SAMPLE_RATE) |
| all_paths.append(p) |
| all_labels.append(0) |
| file_idx += 1 |
| loaded += 1 |
| print(f" Loaded {loaded} real samples") |
| except Exception as e: |
| print(f" MINDS14 failed: {e}") |
|
|
| |
| n_tts = cfg["n_tts_human"] |
| print(f" Generating {n_tts} TTS human samples...") |
| for i in range(n_tts): |
| text = fill(random.choice(cfg["human"]), names, companies) |
| voice, rate, pitch = random.choice(cfg["human_voices"]) |
| p = os.path.join(audio_dir, f"{lang_code}_human_tts_{file_idx:05d}.wav") |
| try: |
| await gen_tts(text, voice, rate, pitch, p) |
| all_paths.append(p); all_labels.append(0); file_idx += 1 |
| except Exception as e: |
| if i % 20 == 0: print(f" tts error: {e}") |
|
|
| |
| n_vm = cfg["n_voicemail"] |
| print(f" Generating {n_vm} voicemail samples...") |
| for i in range(n_vm): |
| text = fill(random.choice(cfg["voicemail"]), names, companies) |
| voice, rate, pitch = random.choice(cfg["voicemail_voices"]) |
| p = os.path.join(audio_dir, f"{lang_code}_voicemail_{file_idx:05d}.wav") |
| try: |
| await gen_tts(text, voice, rate, pitch, p) |
| audio, sr = sf.read(p) |
| beep = generate_tone(random.choice([440, 480, 620, 880]), np.random.uniform(0.3, 0.8), sr, 0.35) |
| fade = min(int(0.01 * sr), len(beep) // 4) |
| beep[:fade] *= np.linspace(0, 1, fade); beep[-fade:] *= np.linspace(1, 0, fade) |
| audio = np.concatenate([audio, silence(0.5, sr), beep, silence(0.7, sr)]) |
| sf.write(p, audio[:int(MAX_LENGTH_S * sr)], sr) |
| all_paths.append(p); all_labels.append(1); file_idx += 1 |
| except Exception as e: |
| if i % 20 == 0: print(f" vm error: {e}") |
|
|
| |
| n_ivr = cfg["n_ivr"] |
| print(f" Generating {n_ivr} IVR samples...") |
| for i in range(n_ivr): |
| text = fill(random.choice(cfg["ivr"]), names, companies) |
| voice, rate, pitch = random.choice(cfg["ivr_voices"]) |
| p = os.path.join(audio_dir, f"{lang_code}_ivr_{file_idx:05d}.wav") |
| try: |
| await gen_tts(text, voice, rate, pitch, p) |
| if random.random() < 0.3: |
| audio, sr = sf.read(p) |
| dtmf = generate_dtmf(random.choice(list('1234567890')), 0.15, sr) |
| audio = np.concatenate([audio, silence(0.8, sr), dtmf, silence(0.3, sr)]) |
| sf.write(p, audio[:int(MAX_LENGTH_S * sr)], sr) |
| all_paths.append(p); all_labels.append(2); file_idx += 1 |
| except Exception as e: |
| if i % 20 == 0: print(f" ivr error: {e}") |
|
|
| |
| n_am = cfg["n_am"] |
| print(f" Generating {n_am} AM samples...") |
| for i in range(n_am): |
| text = fill(random.choice(cfg["am"]), names, companies) |
| voice, rate, pitch = random.choice(cfg["am_voices"]) |
| p = os.path.join(audio_dir, f"{lang_code}_am_{file_idx:05d}.wav") |
| try: |
| await gen_tts(text, voice, rate, pitch, p) |
| audio, sr = sf.read(p) |
| beep = generate_tone(random.choice([1000, 1200, 1400]), np.random.uniform(0.8, 1.5), sr, 0.45) |
| fade = int(0.02 * sr) |
| beep[:fade] *= np.linspace(0, 1, fade); beep[-fade:] *= np.linspace(1, 0, fade) |
| audio = np.concatenate([audio, silence(0.3, sr), beep, silence(1.2, sr)]) |
| sf.write(p, audio[:int(MAX_LENGTH_S * sr)], sr) |
| all_paths.append(p); all_labels.append(3); file_idx += 1 |
| except Exception as e: |
| if i % 20 == 0: print(f" am error: {e}") |
|
|
| print(f" {lang_code.upper()} total: {sum(1 for l in all_labels if True)} cumulative") |
|
|
| |
| print(f"\n{'='*60}") |
| print(f"Multilingual additions: {len(all_paths)} samples") |
| for i, name in enumerate(LABELS): |
| print(f" {name}: {sum(1 for l in all_labels if l == i)}") |
|
|
| |
| print("\nLoading existing dataset...") |
| existing = load_dataset("AbijahKaj/telephony-amd-dataset") |
|
|
| new_ds = Dataset.from_dict({"audio": all_paths, "label": all_labels}) |
| new_ds = new_ds.cast_column("audio", Audio(sampling_rate=SAMPLE_RATE)) |
| new_ds = new_ds.cast_column("label", ClassLabel(names=LABELS)) |
| new_ds = new_ds.shuffle(seed=42) |
| new_splits = new_ds.train_test_split(test_size=0.15, seed=42, stratify_by_column="label") |
|
|
| merged_train = concatenate_datasets([existing["train"], new_splits["train"]]) |
| merged_test = concatenate_datasets([existing["test"], new_splits["test"]]) |
| merged = DatasetDict({"train": merged_train.shuffle(seed=42), "test": merged_test.shuffle(seed=42)}) |
|
|
| print(f"\nMerged dataset:") |
| print(f" Train: {len(merged['train'])}, Test: {len(merged['test'])}") |
| for split in ['train', 'test']: |
| labels = merged[split]['label'] |
| for i, name in enumerate(LABELS): |
| print(f" {split}/{name}: {labels.count(i)}") |
|
|
| print("\nPushing merged dataset to Hub...") |
| merged.push_to_hub("AbijahKaj/telephony-amd-dataset", private=True) |
| print("Done!") |
|
|
| if __name__ == "__main__": |
| asyncio.run(main()) |
|
|
