get phonems

README.md

@@ -104,6 +104,13 @@ python3 decouper_wav_mots_uniquement.py annotation file "mot1" "mot2" "mot3" ...
 python3 matchQueryHS.py /path/to/your/turn/of/speech/npz /path/to/your/queries/npz /path/to/the/output/folder --W1 1 --W2 1
 ```
 
+### • This script get the phonems of the pairs which are mached
+**transcriptionAfterMatching1_fichiers.py :** Output : csv files with transcriptions
+
+```bash
+python3 transcriptionAfterMatching1_fichiers.py /path/to/your/matched/pairs /output/dir textGrid/
+```
+
 # Tools
 
 ### • Counting the number of segments

transcriptionAfterMatching1_fichiers.py

+import os
+import csv
+import re
+from pathlib import Path
+from textgrid import TextGrid
+import argparse
+
+def extract_phonemes_from_textgrid(tg_path, abs_debut, abs_fin, tier_index):
+    tg = TextGrid.fromFile(tg_path)
+    if tier_index >= len(tg):
+        print(f"Tier {tier_index} introuvable dans {tg_path}")
+        return []
+
+    tier = tg[tier_index]
+    phonemes = []
+
+    for interval in tier.intervals:
+        if interval.maxTime > abs_debut and interval.minTime < abs_fin:
+            label = interval.mark.strip()
+            if label:
+                phonemes.append(label)
+
+    return phonemes
+
+def process_csv(input_csv: Path, output_csv: Path, textgrid_dir: Path):
+    with open(input_csv, newline='', encoding='utf-8') as csv_in, \
+         open(output_csv, 'w', newline='', encoding='utf-8') as csv_out:
+
+        reader = csv.reader(csv_in)
+        writer = csv.writer(csv_out)
+
+        header = next(reader)
+        writer.writerow(["file", "phonemes"])
+
+        for row in reader:
+            file_name, t0_str, tn_str, tokens, score = row
+            t0 = float(t0_str)
+            tn = float(tn_str)
+
+            # Extraire start/end du nom de fichier
+            match = re.search(r"_(\d+\.\d+)_(\d+\.\d+)_feature", file_name)
+            if not match:
+                print(f"[ERREUR] Impossible d'extraire start/end depuis : {file_name}")
+                continue
+
+            start_csv = float(match.group(1))
+            end_csv = float(match.group(2))
+
+            abs_debut = start_csv + t0
+            abs_fin = start_csv + tn
+
+            # TextGrid correspondant
+            speaker_id = file_name.split("_")[0].split("-")[0]
+            tg_file = textgrid_dir / f"{speaker_id}.TextGrid"
+            if not tg_file.exists():
+                print(f"[ERREUR] TextGrid introuvable : {tg_file}")
+                continue
+
+            tier_index = 5 if "YBA" in file_name else 2
+            phonemes = extract_phonemes_from_textgrid(tg_file, abs_debut, abs_fin, tier_index)
+            phoneme_str = "-".join(phonemes)
+
+            wav_name = file_name.replace("_feature", ".wav")
+            writer.writerow([wav_name, phoneme_str])
+            print(f"Écrit : {wav_name}, {phoneme_str}")
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Extraire phonèmes depuis TextGrid pour chaque CSV.")
+    parser.add_argument("input_dir", type=Path, help="Dossier contenant les fichiers CSV en entrée")
+    parser.add_argument("output_dir", type=Path, help="Dossier où stocker les CSV transformés")
+    parser.add_argument("textgrid_dir", type=Path, help="Dossier contenant les fichiers TextGrid")
+    args = parser.parse_args()
+
+    args.output_dir.mkdir(parents=True, exist_ok=True)
+
+    csv_files = list(args.input_dir.glob("*.csv"))
+    print(f"Trouvé {len(csv_files)} fichiers CSV dans {args.input_dir}")
+
+    for csv_file in csv_files:
+        output_csv = args.output_dir / csv_file.name
+        print(f"Processing {csv_file} -> {output_csv}")
+        process_csv(csv_file, output_csv, args.textgrid_dir)