amu-cai/polish_presidential_debate-2025

波兰总统辩论2025（BIGOS格式）数据集是一个用于自动语音识别（ASR）评估的语料库，基于BIGOS-V2兼容的重新打包版本。它源自上游语料库directtt/polish_presidential_debate（Apache 2.0许可证），包含195个样本，覆盖13名总统候选人，每名候选人约15个话语，音频来自2025年5月12日的TVP波兰总统辩论。数据集经过重新打包，以静态Parquet格式提供，无需远程代码加载，适用于BIGOS平台评估管道。架构包括音频、参考转录、候选人信息等字段，支持波兰语（pl）环境。原始音频和转录遵循Apache 2.0许可证，而重新打包部分使用CC-BY-NC-4.0许可证。

Polish Presidential Debate 2025 (BIGOS-format) dataset is a corpus for automatic speech recognition (ASR) evaluation, repackaged in a BIGOS-V2-compatible version. It is derived from the upstream corpus directtt/polish_presidential_debate (Apache 2.0 license), containing 195 samples from 13 presidential candidates, with approximately 15 utterances per candidate. The audio originates from the TVP Polish Presidential Debate on 12 May 2025. The dataset is repackaged as static Parquet files to enable loading without remote code, suitable for the BIGOS platform evaluation pipeline. The schema includes fields such as audio, reference transcription, candidate information, and supports the Polish language (pl). The original audio and transcripts are under Apache 2.0 license, while the repackaging curation uses CC-BY-NC-4.0 license.