riigikogu-audio-stenograms-2018-2025

Riigikogu Stenograms 2018-2025数据集包含爱沙尼亚议会（Riigikogu）2017年底至2025年5月期间的会议速记稿（近似转录文本）及对应音频。这些转录文本并非逐字记录，而是经过编辑以提高可读性和语法正确性，并提供句子级别的起止时间标记（相对于对应音频文件）。数据集以Transcriber XML（.trs）格式呈现，包含发言者姓名和议题名称（如"Tulumaksuseaduse muutmise seaduse eelnõu esimene lugemine"）。数据源自Riigikogu开放API和YouTube，通过基于多语言Wav2Vec2 CTC语音模型的强制对齐流程处理，确保99%以上的句子时间戳精确到秒级。该数据集共包含1001个会议录音，总时长约3084小时。The Riigikogu Stenograms 2018-2025 dataset contains parliamentary stenographic transcripts (approximate transcription texts) and corresponding audio recordings of the Estonian parliament (Riigikogu) covering the period from late 2017 to May 2025. These transcripts are not verbatim records; instead, they have been edited to enhance readability and grammatical correctness, and include sentence-level start and end timestamps relative to their corresponding audio files. The dataset is provided in Transcriber XML (.trs) format, which includes speaker names and topic names (e.g., "Tulumaksuseaduse muutmise seaduse eelnõu esimene lugemine"). The data is sourced from the Riigikogu open API and YouTube, and processed through a forced alignment pipeline based on the multilingual Wav2Vec2 CTC speech model, ensuring that over 99% of sentence timestamps are accurate to the second level. In total, the dataset includes 1,001 meeting recordings with a combined duration of approximately 3,084 hours.