annajuliaasf/Audio-Transcription-Models-Comparison-PT-BR

这是一个专注于比较现代语音转文本（STT）模型性能的数据集，特别针对巴西葡萄牙语（PT-BR）。数据集旨在评估不同AI模型在巴西实际使用场景中的表现，包括处理区域方言、非正式语言和数字实体等挑战。所有音频样本都经过标准化处理，转换为无损WAV格式并统一为16kHz采样率，以确保比较的公平性。数据集还包含了由OpenAI Whisper、OpenAI GPT-4o-mini-transcribe和Google Gemini-2.0-Flash-Exp等模型生成的转录结果。

A dataset dedicated to comparing the performance of modern Speech-to-Text (STT) models, focusing exclusively on Brazilian Portuguese. This dataset was created to store and compare transcription results from different Artificial Intelligence models in challenging scenarios, covering regionalism, informality and disfluency, and numeric entities. All audio samples were converted to the uncompressed WAV format and standardized to 16kHz to ensure fairness in comparison. The dataset includes transcriptions generated by models such as OpenAI Whisper, OpenAI GPT-4o-mini-transcribe, and Google Gemini-2.0-Flash-Exp.