Convert dataset to JSONL format for better Hugging Face compatibility

- Converted CSV dataset to JSONL format with structured audio metadata
- Updated all Python scripts to work with JSONL instead of CSV
- Added proper audio sampling rate metadata (22050 Hz)
- Removed old CSV file in favor of JSONL
- Ensures word and IPA columns are properly displayed in Dataset Viewer

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

convert_to_jsonl.py

@@ -0,0 +1,45 @@
+#!/usr/bin/env python3
+
+import csv
+import json
+from pathlib import Path
+
+
+def convert_csv_to_jsonl(csv_file="vietnamese_ipa_dataset.csv", jsonl_file="vietnamese_ipa_dataset.jsonl"):
+    """Convert CSV dataset to JSONL format"""
+
+    records = []
+
+    # Read CSV file
+    with open(csv_file, "r", encoding="utf-8") as f:
+        reader = csv.DictReader(f)
+        for row in reader:
+            # Create record with proper structure
+            record = {
+                "word": row["word"],
+                "ipa": row["ipa"],
+                "audio": {
+                    "path": row["audio_file"],
+                    "sampling_rate": 22050  # Standard sampling rate for gTTS
+                }
+            }
+            records.append(record)
+
+    # Write JSONL file
+    with open(jsonl_file, "w", encoding="utf-8") as f:
+        for record in records:
+            f.write(json.dumps(record, ensure_ascii=False) + "\n")
+
+    print(f"Converted {len(records)} records from CSV to JSONL")
+    print(f"Output saved to: {jsonl_file}")
+
+    # Display first 5 records
+    print("\nFirst 5 records:")
+    for i, record in enumerate(records[:5], 1):
+        print(f"{i}. {record['word']} -> {record['ipa']}")
+
+    return records
+
+
+if __name__ == "__main__":
+    convert_csv_to_jsonl()

example_usage.py

@@ -1,11 +1,17 @@
 #!/usr/bin/env python3
 
+import json
 import pandas as pd
 import os
 
 
-def load_dataset(csv_file="vietnamese_ipa_dataset.csv"):
-    df = pd.read_csv(csv_file)
+def load_dataset(jsonl_file="vietnamese_ipa_dataset.jsonl"):
+    records = []
+    with open(jsonl_file, "r", encoding="utf-8") as f:
+        for line in f:
+            records.append(json.loads(line))
+
+    df = pd.DataFrame(records)
     return df
 
 
@@ -37,8 +43,9 @@ def main():
     print("-" * 50)
 
     for idx, row in df.head(10).iterrows():
+        audio_path = row["audio"]["path"] if isinstance(row["audio"], dict) else row["audio"]
         print(
-            f"{idx + 1}. {row['word']:<15} IPA: {row['ipa']:<25} Audio: {row['audio_file']}"
+            f"{idx + 1}. {row['word']:<15} IPA: {row['ipa']:<25} Audio: {audio_path}"
         )
 
     print("\n" + "=" * 50)
@@ -56,13 +63,14 @@ def main():
         if results:
             print(f"\nFound {len(results)} result(s):")
             for i, result in enumerate(results, 1):
+                audio_path = result["audio"]["path"] if isinstance(result["audio"], dict) else result["audio"]
                 print(f"\n  {i}. Word: {result['word']}")
                 print(f"     IPA: {result['ipa']}")
-                print(f"     Audio: {result['audio_file']}")
+                print(f"     Audio: {audio_path}")
 
                 play_choice = input("     Play audio? (y/n): ").strip().lower()
                 if play_choice == "y":
-                    if play_audio(result["audio_file"]):
+                    if play_audio(audio_path):
                         print("     ✓ Playing audio...")
                     else:
                         print("     ✗ Audio file not found")

generate_audio.py

@@ -1,6 +1,6 @@
 #!/usr/bin/env python3
 
-import csv
+import json
 from pathlib import Path
 from gtts import gTTS
 import pandas as pd
@@ -23,25 +23,30 @@ def generate_mp3_from_text(text, output_path, lang="vi"):
         return False
 
 
-def process_dataset(csv_file="vietnamese_ipa_dataset.csv"):
+def process_dataset(jsonl_file="vietnamese_ipa_dataset.jsonl"):
     create_audio_directory()
 
-    df = pd.read_csv(csv_file)
+    # Read JSONL file
+    records = []
+    with open(jsonl_file, "r", encoding="utf-8") as f:
+        for line in f:
+            records.append(json.loads(line))
+
+    df = pd.DataFrame(records)
 
     total = len(df)
     success = 0
     failed = []
 
-    print(f"Processing {total} words from {csv_file}...")
+    print(f"Processing {total} words from {jsonl_file}...")
     print("-" * 50)
 
     for idx, row in df.iterrows():
         word = row["word"]
         ipa = row["ipa"]
-        audio_file = row["audio_file"]
+        audio_path = row["audio"]["path"] if isinstance(row["audio"], dict) else row["audio"]
 
-        audio_path = Path(audio_file)
-        full_path = audio_path
+        full_path = Path(audio_path)
 
         print(f"[{idx + 1}/{total}] Generating: {word} [{ipa}]")
 
@@ -61,12 +66,22 @@ def process_dataset(csv_file="vietnamese_ipa_dataset.csv"):
     return success, failed
 
 
-def add_new_word(word, ipa, csv_file="vietnamese_ipa_dataset.csv"):
+def add_new_word(word, ipa, jsonl_file="vietnamese_ipa_dataset.jsonl"):
     audio_filename = f"audio/{word.replace(' ', '_')}.mp3"
 
-    with open(csv_file, "a", newline="", encoding="utf-8") as f:
-        writer = csv.writer(f)
-        writer.writerow([word, ipa, audio_filename])
+    # Create new record
+    record = {
+        "word": word,
+        "ipa": ipa,
+        "audio": {
+            "path": audio_filename,
+            "sampling_rate": 22050
+        }
+    }
+
+    # Append to JSONL file
+    with open(jsonl_file, "a", encoding="utf-8") as f:
+        f.write(json.dumps(record, ensure_ascii=False) + "\n")
 
     create_audio_directory()
     if generate_mp3_from_text(word, audio_filename):
@@ -77,15 +92,22 @@ def add_new_word(word, ipa, csv_file="vietnamese_ipa_dataset.csv"):
         return False
 
 
-def display_dataset(csv_file="vietnamese_ipa_dataset.csv"):
-    df = pd.read_csv(csv_file)
+def display_dataset(jsonl_file="vietnamese_ipa_dataset.jsonl"):
+    # Read JSONL file
+    records = []
+    with open(jsonl_file, "r", encoding="utf-8") as f:
+        for line in f:
+            records.append(json.loads(line))
+
+    df = pd.DataFrame(records)
     print("\nVietnamese IPA Dataset:")
     print("-" * 70)
     print(f"{'Word':<20} {'IPA':<30} {'Audio File':<20}")
     print("-" * 70)
 
     for _, row in df.iterrows():
-        print(f"{row['word']:<20} {row['ipa']:<30} {row['audio_file']:<20}")
+        audio_path = row["audio"]["path"] if isinstance(row["audio"], dict) else row["audio"]
+        print(f"{row['word']:<20} {row['ipa']:<30} {audio_path:<20}")
 
     print("-" * 70)
     print(f"Total entries: {len(df)}")

vietnamese_ipa_dataset.csv

@@ -1,51 +0,0 @@
-word,ipa,audio_file
-xin chào,sin˧˧ caːw˨˩,audio/xin_chao.mp3
-cảm ơn,kaːm˧˩ ʔəːn˧˧,audio/cam_on.mp3
-Việt Nam,viət˧˩ˀ naːm˧˧,audio/viet_nam.mp3
-nước,nɨək˧˥,audio/nuoc.mp3
-ăn,ăn˧˧,audio/an.mp3
-uống,ʔuəŋ˧˥,audio/uong.mp3
-nhà,ɲaː˨˩,audio/nha.mp3
-trường,ʈɨəŋ˨˩,audio/truong.mp3
-bạn,ɓaːn˧˩ˀ,audio/ban.mp3
-yêu,iəw˧˧,audio/yeu.mp3
-đẹp,ɗɛp˧˩ˀ,audio/dep.mp3
-sách,saːk˧˥,audio/sach.mp3
-hoa,hwaː˧˧,audio/hoa.mp3
-mẹ,mɛ˧˩ˀ,audio/me.mp3
-ba,ɓaː˧˧,audio/ba.mp3
-con,kɔn˧˧,audio/con.mp3
-chó,cɔ˧˥,audio/cho.mp3
-mèo,mɛːw˨˩,audio/meo.mp3
-cây,kəj˧˧,audio/cay.mp3
-lá,laː˧˥,audio/la.mp3
-trời,ʈəːj˨˩,audio/troi.mp3
-mưa,mɨə˧˧,audio/mua.mp3
-gió,zɔ˧˥,audio/gio.mp3
-nắng,năŋ˧˥,audio/nang.mp3
-đường,ɗɨəŋ˨˩,audio/duong.mp3
-phố,fo˧˥,audio/pho.mp3
-làm,laːm˨˩,audio/lam.mp3
-việc,viək˧˩ˀ,audio/viec.mp3
-học,hɔk˧˩ˀ,audio/hoc.mp3
-sinh,siɲ˧˧,audio/sinh.mp3
-thầy,thəj˨˩,audio/thay.mp3
-cô,ko˧˧,audio/co.mp3
-tôi,toj˧˧,audio/toi.mp3
-anh,aːɲ˧˧,audio/anh.mp3
-em,ɛm˧˧,audio/em.mp3
-chị,ci˧˩ˀ,audio/chi.mp3
-ông,ʔoŋ˧˧,audio/ong.mp3
-bà,ɓaː˨˩,audio/ba_tone.mp3
-năm,năm˧˧,audio/nam.mp3
-tháng,thaːŋ˧˥,audio/thang.mp3
-ngày,ŋaːj˨˩,audio/ngay.mp3
-sáng,saːŋ˧˥,audio/sang.mp3
-trưa,ʈɨə˧˧,audio/trua.mp3
-chiều,ciəw˨˩,audio/chieu.mp3
-tối,toj˧˥,audio/toi_tone.mp3
-đêm,ɗem˧˧,audio/dem.mp3
-một,mot˧˩ˀ,audio/mot.mp3
-hai,haːj˧˧,audio/hai.mp3
-ba,ɓaː˧˧,audio/ba_number.mp3
-bốn,ɓon˧˥,audio/bon.mp3

vietnamese_ipa_dataset.jsonl

@@ -0,0 +1,50 @@
+{"word": "xin chào", "ipa": "sin˧˧ caːw˨˩", "audio": {"path": "audio/xin_chao.mp3", "sampling_rate": 22050}}
+{"word": "cảm ơn", "ipa": "kaːm˧˩ ʔəːn˧˧", "audio": {"path": "audio/cam_on.mp3", "sampling_rate": 22050}}
+{"word": "Việt Nam", "ipa": "viət˧˩ˀ naːm˧˧", "audio": {"path": "audio/viet_nam.mp3", "sampling_rate": 22050}}
+{"word": "nước", "ipa": "nɨək˧˥", "audio": {"path": "audio/nuoc.mp3", "sampling_rate": 22050}}
+{"word": "ăn", "ipa": "ăn˧˧", "audio": {"path": "audio/an.mp3", "sampling_rate": 22050}}
+{"word": "uống", "ipa": "ʔuəŋ˧˥", "audio": {"path": "audio/uong.mp3", "sampling_rate": 22050}}
+{"word": "nhà", "ipa": "ɲaː˨˩", "audio": {"path": "audio/nha.mp3", "sampling_rate": 22050}}
+{"word": "trường", "ipa": "ʈɨəŋ˨˩", "audio": {"path": "audio/truong.mp3", "sampling_rate": 22050}}
+{"word": "bạn", "ipa": "ɓaːn˧˩ˀ", "audio": {"path": "audio/ban.mp3", "sampling_rate": 22050}}
+{"word": "yêu", "ipa": "iəw˧˧", "audio": {"path": "audio/yeu.mp3", "sampling_rate": 22050}}
+{"word": "đẹp", "ipa": "ɗɛp˧˩ˀ", "audio": {"path": "audio/dep.mp3", "sampling_rate": 22050}}
+{"word": "sách", "ipa": "saːk˧˥", "audio": {"path": "audio/sach.mp3", "sampling_rate": 22050}}
+{"word": "hoa", "ipa": "hwaː˧˧", "audio": {"path": "audio/hoa.mp3", "sampling_rate": 22050}}
+{"word": "mẹ", "ipa": "mɛ˧˩ˀ", "audio": {"path": "audio/me.mp3", "sampling_rate": 22050}}
+{"word": "ba", "ipa": "ɓaː˧˧", "audio": {"path": "audio/ba.mp3", "sampling_rate": 22050}}
+{"word": "con", "ipa": "kɔn˧˧", "audio": {"path": "audio/con.mp3", "sampling_rate": 22050}}
+{"word": "chó", "ipa": "cɔ˧˥", "audio": {"path": "audio/cho.mp3", "sampling_rate": 22050}}
+{"word": "mèo", "ipa": "mɛːw˨˩", "audio": {"path": "audio/meo.mp3", "sampling_rate": 22050}}
+{"word": "cây", "ipa": "kəj˧˧", "audio": {"path": "audio/cay.mp3", "sampling_rate": 22050}}
+{"word": "lá", "ipa": "laː˧˥", "audio": {"path": "audio/la.mp3", "sampling_rate": 22050}}
+{"word": "trời", "ipa": "ʈəːj˨˩", "audio": {"path": "audio/troi.mp3", "sampling_rate": 22050}}
+{"word": "mưa", "ipa": "mɨə˧˧", "audio": {"path": "audio/mua.mp3", "sampling_rate": 22050}}
+{"word": "gió", "ipa": "zɔ˧˥", "audio": {"path": "audio/gio.mp3", "sampling_rate": 22050}}
+{"word": "nắng", "ipa": "năŋ˧˥", "audio": {"path": "audio/nang.mp3", "sampling_rate": 22050}}
+{"word": "đường", "ipa": "ɗɨəŋ˨˩", "audio": {"path": "audio/duong.mp3", "sampling_rate": 22050}}
+{"word": "phố", "ipa": "fo˧˥", "audio": {"path": "audio/pho.mp3", "sampling_rate": 22050}}
+{"word": "làm", "ipa": "laːm˨˩", "audio": {"path": "audio/lam.mp3", "sampling_rate": 22050}}
+{"word": "việc", "ipa": "viək˧˩ˀ", "audio": {"path": "audio/viec.mp3", "sampling_rate": 22050}}
+{"word": "học", "ipa": "hɔk˧˩ˀ", "audio": {"path": "audio/hoc.mp3", "sampling_rate": 22050}}
+{"word": "sinh", "ipa": "siɲ˧˧", "audio": {"path": "audio/sinh.mp3", "sampling_rate": 22050}}
+{"word": "thầy", "ipa": "thəj˨˩", "audio": {"path": "audio/thay.mp3", "sampling_rate": 22050}}
+{"word": "cô", "ipa": "ko˧˧", "audio": {"path": "audio/co.mp3", "sampling_rate": 22050}}
+{"word": "tôi", "ipa": "toj˧˧", "audio": {"path": "audio/toi.mp3", "sampling_rate": 22050}}
+{"word": "anh", "ipa": "aːɲ˧˧", "audio": {"path": "audio/anh.mp3", "sampling_rate": 22050}}
+{"word": "em", "ipa": "ɛm˧˧", "audio": {"path": "audio/em.mp3", "sampling_rate": 22050}}
+{"word": "chị", "ipa": "ci˧˩ˀ", "audio": {"path": "audio/chi.mp3", "sampling_rate": 22050}}
+{"word": "ông", "ipa": "ʔoŋ˧˧", "audio": {"path": "audio/ong.mp3", "sampling_rate": 22050}}
+{"word": "bà", "ipa": "ɓaː˨˩", "audio": {"path": "audio/ba_tone.mp3", "sampling_rate": 22050}}
+{"word": "năm", "ipa": "năm˧˧", "audio": {"path": "audio/nam.mp3", "sampling_rate": 22050}}
+{"word": "tháng", "ipa": "thaːŋ˧˥", "audio": {"path": "audio/thang.mp3", "sampling_rate": 22050}}
+{"word": "ngày", "ipa": "ŋaːj˨˩", "audio": {"path": "audio/ngay.mp3", "sampling_rate": 22050}}
+{"word": "sáng", "ipa": "saːŋ˧˥", "audio": {"path": "audio/sang.mp3", "sampling_rate": 22050}}
+{"word": "trưa", "ipa": "ʈɨə˧˧", "audio": {"path": "audio/trua.mp3", "sampling_rate": 22050}}
+{"word": "chiều", "ipa": "ciəw˨˩", "audio": {"path": "audio/chieu.mp3", "sampling_rate": 22050}}
+{"word": "tối", "ipa": "toj˧˥", "audio": {"path": "audio/toi_tone.mp3", "sampling_rate": 22050}}
+{"word": "đêm", "ipa": "ɗem˧˧", "audio": {"path": "audio/dem.mp3", "sampling_rate": 22050}}
+{"word": "một", "ipa": "mot˧˩ˀ", "audio": {"path": "audio/mot.mp3", "sampling_rate": 22050}}
+{"word": "hai", "ipa": "haːj˧˧", "audio": {"path": "audio/hai.mp3", "sampling_rate": 22050}}
+{"word": "ba", "ipa": "ɓaː˧˧", "audio": {"path": "audio/ba_number.mp3", "sampling_rate": 22050}}
+{"word": "bốn", "ipa": "ɓon˧˥", "audio": {"path": "audio/bon.mp3", "sampling_rate": 22050}}