juanjucm/OpenSLR-SpeechT-GL-EN

OpenSLR-SpeechT-GL-EN数据集是一个用于加利亚语到英语语音翻译任务的数据集。该数据集包含了约10小时20分钟的加利亚语男女音频及其文本转录和相应的英语翻译。数据集结构包括训练集、验证集和测试集，每个集都包含音频、加利亚语文本、英语文本和说话者信息。数据预处理步骤包括去除重复的转录实例，并使用GPT4-o模型进行机器翻译生成英语文本。翻译质量通过质量估计模型进行评估，平均得分为0.75。

OpenSLR-SpeechT-GL-EN is a Galician-to-English dataset for Speech Translation task. It contains approximately 10 hours and 20 minutes of Galician male and female audio along with its text transcriptions and the corresponding English translations. The English translations in the dataset are created from scratch through a Machine Translation step applied to the Galician transcriptions. The dataset structure includes a training set with three features: audio, Galician text, and English text, totaling 5587 rows.