You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

Stage 3 資料集組裝與切分統計報告

📊 全語料總覽 (Global Summary)

  • 有效語言總數: 3 個語言
  • 總語音樣本數: 14,903 筆
  • 總音訊總時長: 28.72 小時 (103,379.5 秒)
  • 切分策略: Excel 原生工作表切分 (use_sheet_split=True)
  • 剔除重複 (Train/Eval Overlap Excluded): 22 筆 (已從 train 排除以避免資料洩漏)
  • Train 集規模: 14,843 筆 (28.60 小時)
  • Eval 集規模: 60 筆 (0.11 小時, 佔比: 0.4%)

📋 各語言分切明細 (By Language)

語言別 (Language) ISO 代碼 Train 筆數 Train 時長 (hr) Eval 筆數 Eval 時長 (hr) 總筆數 總時長 (hr) Eval 佔比
太魯閣語 trv-x-truku 7,629 8.83 20 0.03 7,649 8.86 0.3%
德固達雅賽德克語 trv-x-tgdy 3,847 10.32 20 0.04 3,867 10.36 0.5%
秀姑巒阿美語 ami-x-skl 3,367 9.45 20 0.04 3,387 9.49 0.6%
總計 (Total) - 14,843 28.60 60 0.11 14,903 28.72 0.4%
Downloads last month
112