MCE 2018数据集

MCE 2018数据集是由麻省理工学院计算机科学与人工智能实验室创建，用于评估当前语音技术在多目标说话人检测与识别方面的能力。该数据集包含来自呼叫中心客户与代理对话的录音，每个对话由一个ivector表示。数据集分为训练集和开发集，包含3631个黑名单说话人，每个说话人在训练集中出现3次，在开发集中出现1次。数据集的应用领域是解决在实际电话对话中识别黑名单说话人的问题，旨在提高语音识别系统的准确性和鲁棒性。

The MCE 2018 dataset was created by the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) to evaluate the capabilities of current speech technologies in multi-target speaker detection and recognition. This dataset includes audio recordings of conversations between call center customers and agents, with each conversation represented by an ivector. The dataset is divided into a training set and a development set, containing 3631 blacklisted speakers. Each speaker appears three times in the training set and once in the development set. The dataset is intended to address the problem of identifying blacklisted speakers in real-world telephone conversations, aiming to improve the accuracy and robustness of speech recognition systems.