Active learning method and device for labeling platform

The invention discloses an active learning method and device for a labeling platform. The method comprises the following steps of training a text model by utilizing the labeled data; predicting the labels of the unlabeled data by utilizing the trained text model, and calculating the uncertainty of each piece of unlabeled data; based on the uncertainty of each piece of unlabeled data, determining the labeling sequence of the unlabeled data; and recommending the unlabeled data and the labels thereof to the user according to the labeling sequence of the unlabeled data. 本申请公开一种用于标注平台的主动学习方法及装置，所述方法包括：利用已经标注数据训练文本模型；利用已经训练的文本模型预测未标注数据的标签，并计算每条未标注数据的不确定性；基于每条未标注数据的不确定性，确定未标注数据的标注顺序；按照未标注数据的标注顺序，将未标注数据及其标签推荐给用户。