Missing audio data for questions and answers in `zho_Hans` subset

[YoungSeng]

Hi there,

I am working with the Belebele-speech dataset and noticed an issue specifically with the zho_Hans (Chinese Simplified) subset.

While the passage audio seems to load correctly, the audio data for the questions and answers appears to be missing (None). This works fine for other languages I tested (e.g., fra_Latn), where these columns contain valid audio data.

Observed Behavior:
When loading zho_Hans.parquet, the following columns are empty:

• question_audio
• answer_1_audio
• answer_2_audio
• answer_3_audio
• answer_4_audio

Reproduction Steps:

import polars as pl

# Path to the zho_Hans parquet file
parquet_path = "path/to/lang=zho_Hans/zho_Hans.parquet"
df = pl.read_parquet(parquet_path)

# Check the first row
row = df.head(1).to_dicts()[0]

print(f"Language: {row.get('lang', 'zho_Hans')}")
print(f"Question Audio: {row['question_audio']}") 
# Output is None (or empty list)
print(f"Answer 1 Audio: {row['answer_1_audio']}") 
# Output is None