add cv to accent classifier

audio_classification_scripts/run_mms_lid_with_cv.sh

+#!/usr/bin/env bash
+
+python run_audio_classification.py \
+    --model_name_or_path "facebook/mms-lid-126" \
+    --train_dataset_name "stable-speech/concatenated-normalized-accent-dataset+stable-speech/concatenated-common-voice-15-accented" \
+    --train_dataset_config_name "default+default" \
+    --train_split_name "train+train" \
+    --train_label_column_name "labels+labels" \
+    --eval_dataset_name "stable-speech/concatenated-normalized-accent-dataset" \
+    --eval_dataset_config_name "default" \
+    --eval_split_name "test" \
+    --eval_label_column_name "labels" \
+    --output_dir "./" \
+    --do_train \
+    --do_eval \
+    --overwrite_output_dir \
+    --remove_unused_columns False \
+    --fp16 \
+    --fp16_full_eval \
+    --learning_rate 1e-4 \
+    --lr_scheduler_type "constant_with_warmup" \
+    --max_length_seconds 20 \
+    --min_length_seconds 5 \
+    --attention_mask \
+    --warmup_steps 100 \
+    --max_steps 5000 \
+    --per_device_train_batch_size 32 \
+    --per_device_eval_batch_size 32 \
+    --preprocessing_num_workers 4 \
+    --dataloader_num_workers 4 \
+    --logging_strategy "steps" \
+    --logging_steps 10 \
+    --evaluation_strategy "steps" \
+    --eval_steps 1000 \
+    --save_strategy "no" \
+    --save_steps 5000 \
+    --filter_threshold 0.01 \
+    --freeze_base_model False \
+    --push_to_hub False \
+    --trust_remote_code

dataset_concatenation_scripts/run_dataset_concatenation_cv.sh

+#!/usr/bin/env bash
+
+python run_dataset_concatenation.py \
+    --dataset_name "stable-speech/common_voice_15_0_accented" \
+    --dataset_config_name "en" \
+    --dataset_split_name "train" \
+    --label_column_name "accent" \
+    --text_column_name "sentence" \
+    --speaker_column_name "client_id" \
+    --batch_size 250 \
+    --preprocessing_num_workers 4 \
+    --output_dir "./concatenated-dataset-cv"
+
+python run_dataset_concatenation.py \
+    --dataset_name "stable-speech/common_voice_15_0_accented" \
+    --dataset_config_name "en" \
+    --dataset_split_name "test" \
+    --label_column_name "accent" \
+    --text_column_name "sentence" \
+    --speaker_column_name "client_id" \
+    --batch_size 250 \
+    --preprocessing_num_workers 4 \
+    --output_dir "./concatenated-dataset-cv-test"

edacc/prepare_edacc.py

@@ -76,14 +76,17 @@ def main():
     accent_dataset = load_dataset("sanchit-gandhi/edacc_accents", split="train")
 
     def format_dataset(batch):
-        batch["speaker_id"]  = batch["Final-Participant_ID"].replace("EAEC", "EDACC").replace("P1", "-A").replace("P2", "-B")
+        batch["speaker_id"] = (
+            batch["Final-Participant_ID"].replace("EAEC", "EDACC").replace("P1", "-A").replace("P2", "-B")
+        )
         return batch
 
     accent_dataset = accent_dataset.map(format_dataset, remove_columns=["Final-Participant_ID"])
 
     # 2. Clean accents for each speaker
     linguistic_background_clean = {
-        participant: accent.strip() for participant, accent in zip(accent_dataset["speaker_id"], accent_dataset["English_Variety"])
+        participant: accent.strip()
+        for participant, accent in zip(accent_dataset["speaker_id"], accent_dataset["English_Variety"])
     }
     linguistic_variety = {
         participant: l1.strip() for participant, l1 in zip(accent_dataset["speaker_id"], accent_dataset["L1_Variety"])

run_audio_classification.py

@@ -57,21 +57,22 @@ def random_subsample(wav: np.ndarray, max_length: float, sample_rate: int = 1600
     random_offset = randint(0, len(wav) - sample_length - 1)
     return wav[random_offset : random_offset + sample_length]
 
+
 def deterministic_subsample(wav: np.ndarray, max_length: float, sample_rate: int = 16000) -> np.ndarray:
     """Take first `max_length` seconds from the input audio"""
     sample_length = int(round(sample_rate * max_length))
     if len(wav) <= sample_length:
         return wav
-    return wav[0 : sample_length]
+    return wav[0:sample_length]
 
 
 ACCENT_MAPPING = {
     "British": "English",
-    "Canadian": "American",
+    # "Canadian": "American",
     "Northern irish": "Irish",
-    "New zealand": "Australian",
+    # "New zealand": "Australian",
     "Pakistani": "Indian",
-    "Mainstream us english": "American",
+    "Mainstream u s english": "American",
     "Southern british english": "English",
     "Indian english": "Indian",
     "Scottish english": "Scottish",
@@ -83,12 +84,152 @@ ACCENT_MAPPING = {
     "Indonesian english": "Indonesian",
     "South african english": "South african",
     "Irish english": "Irish",
-    "Latin": "Latin American",
+    "Latin": "Latin american",
     "European": "Unknown",  # Too general
-    "Eastern european": "Eastern european", # TODO(SG): keep for now, but maybe remove later
+    "Eastern european": "Eastern european",  # TODO(SG): keep for now, but maybe remove later as too general
+    "Bangladeshi": "Indian",
+    "England": "English",
+    "India": "Indian",
+    "Afrikaans": "South african",
+    "California": "American",
+    "Nepali": "Indian",
+    "New york city": "American",
+    "New jerseyan": "American",
+    "Northumbrian british english": "English",
+    "Nottinghamshire,east midlands": "English",
+    "Southern african": "South african",
+    "United states english": "American",
+    "West indies": "Jamaican",
+    "2nd language": "Unknown",  # Too vague
+    "A savage texas gentleman": "American",
+    "A variety of texan english with some german influence that has undergone the cot-caught merger": "American",
+    "A'lo": "Unknown",  # Unclear
+    "Academic southern english,england english": "English",
+    "Argentinian english": "Latin american",
+    "Austrian": "German",
+    "Bangladesh,india and south asia (india, pakistan, sri lanka)": "Indian",
+    "Brazillian accent": "Brazilian",
+    "British accent": "English",
+    "Caribbean canadian": "Unknown",  # Specific combination not listed
+    "Colombian accent": "Latin american",
+    "Czech accent": "Czech",
+    "East african khoja": "Unknown",  # Specific community
+    "East indian": "Indian",
+    "East london": "English",
+    "England,london,academic": "English",
+    "Filipino": "Unknown",  # Unique blend
+    "Fluent,e sl,european": "Unknown",  # Too vague
+    "Generic european": "Unknown",  # Too vague
+    "Georgian english": "Unknown",  # No direct match
+    "Ghanaian english accent,african regular reader": "Unknown",  # Specific category not listed
+    "Haitian creole": "Unknown",  # Unique blend
+    "Hispanic": "Latin american",
+    "Hispanic/latino": "Latin american",
+    "Hong kong english": "Chinese",
+    "Hong kong english,scottish english": "Chinese",
+    "Hunglish": "Hungarian",
+    "I think mine accent is influenced by indian accent ,yes please. ,india and south asia (india, pakistan, sri lanka)": "Indian",
+    "I was born in england and have lived in australia, canada and france.": "English",
+    "International english,united states english,australian english": "American",
+    "Israeli": "Unknown",  # No direct match
+    "Israeli english": "Unknown",  # No direct match
+    "Javanese,indonesian english,malaysian english": "Indonesian",
+    "Kazakhstan english": "Unknown",  # No direct match
+    "Kiwi": "New zealand",  # Could be generalised to Australian
+    "Latin america,united states english": "Latin american",
+    "Latin american accent": "Latin american",
+    "Latin english": "Unknown",  # Too vague
+    "Latino": "Latin american",
+    "Latvian": "Latvian",  # Note: added new
+    "Little latino,united states english,second language": "Latin american",
+    "Liverpool english,lancashire english,england english": "English",
+    "Liverpudlian english": "English",
+    "Malaysian english": "Malaysian",  # Note: added new
+    "Mexican accent": "Latin american",
+    "Mid-atlantic united states english,philadelphia, pennsylvania, united states english,united states english,philadelphia style united states english": "American",
+    "Mid-atlantic,england english,united states english": "American",
+    "Midatlantic,england english": "American",
+    "Midwestern states (michigan),united states english": "American",
+    "Mild northern england english": "English",
+    "Minor french accent": "French",
+    "Mix of american and british ,native polish": "Polish",
+    "Mix of american and british accent": "Unknown",  # Combination not clearly mapped
+    "Mostly american with some british and australian inflections": "Unknown",  # Combination not clearly mapped
+    "My accent is influenced by the phones of all letters within a sentence.,southern african (south africa, zimbabwe, namibia)": "South african",
+    "New zealand english": "New Zealand English",
+    "Nigeria english": "Nigerian",  # Note: added new
+    "Non native speaker from france": "French",
+    "Non-native": "Unknown",  # Too vague
+    "Non-native,german accent": "German",
+    "North european english": "Unknown",  # Too broad
+    "Norwegian": "Norwegian",  # Note: added new
+    "Ontario,canadian english": "Canadian",  # Note: added new
+    "Polish english": "Polish",
+    "Rhode island new england accent": "American",
+    "Singaporean english": "Singaporean",  # Note: added new
+    "Slavic": "Eastern european",
+    "Slighty southern affected by decades in the midwest, 4 years in spain and germany, speak some german, spanish, polish. have lived in nine states.": "Unknown",  # Complex blend
+    "South african": "South african",
+    "South atlantic (falkland islands, saint helena)": "Unknown",  # Specific regions not listed
+    "South australia": "Australian",
+    "South indian": "Indian",
+    "Southern drawl": "American",
+    "Southern texas accent,united states english": "American",
+    "Southern united states,united states english": "American",
+    "Spanish bilingual": "Spanish",
+    "Spanish,foreign,non-native": "Spanish",
+    "Strong latvian accent": "Latvian",
+    "Swedish accent": "Swedish",  # Note: added new
+    "Transnational englishes blend": "Unknown",  # Too vague
+    "U.k. english": "English",
+    "Very slight russian accent,standard american english,boston influence": "American",
+    "Welsh english": "Welsh",
+    "West african": "Unknown",  # No specific West African category
+    "West indian": "Unknown",  # Caribbean, but no specific match
+    "Western europe": "Unknown",  # Too broad
+    "With heavy cantonese accent": "Chinese",
 }
 
 
+STARTS_WITH = [
+    "Afrikaans",
+    "American",
+    "Australian",
+    "Bangladeshi",
+    "Canadian",
+    "Chinese",
+    "Dutch",
+    "Eastern European",
+    "European",
+    "England",
+    "English",
+    "German",
+    "Filipino",
+    "India",
+    "Irish" "Israeli",
+    "Italian",
+    "Japanese",
+    "Kenyan",
+    "Northern Irish",
+    "New Zealand",
+    "Nigerian",
+    "Malaysian",
+    "Russian",
+    "Scottish",
+    "Singaporean",
+    "Slavic",
+    "South African",
+    "Southern African",
+    "Swedish",
+    "Swiss",
+    "United States English",
+    "West Indies",
+    "french",
+    "polish",
+    "serbian",
+]
+
+
 def preprocess_labels(label: str) -> str:
     """Apply pre-processing formatting to the accent labels"""
     if "_" in label:
@@ -96,7 +237,10 @@ def preprocess_labels(label: str) -> str:
         language_code = label.split("_")[-1]
         label = LANGUAGES[language_code]
     # VCTK labels for two words are concatenated into one (NewZeleand-> New Zealand)
-    label = re.sub(r"(\w)([A-Z])", r"\1 \2", label)
+    label = re.sub(r"(\w)([A-Z])", r"\1 \2", label).strip()
+    for prefix in STARTS_WITH:
+        if label.startswith(prefix):
+            label = prefix
     # convert Whisper language code (polish) to capitalised (Polish)
     label = label.capitalize()
     if label in ACCENT_MAPPING: