基于多任务的多标签文本分类

Multi-Label Text Classification Based on Multi-Task

在自然语言处理任务中,多标签文本分类是一项具有挑战性的任务。传统的文本分类任务只需要将每条样本中的特征与一个特定类别关联起来就可以得到很好的预测结果。多标签文本分类任务中的每个样本对应了一个或多个标签,每个样本中的特征难以清晰地映射到对应的各个标签上。针对多标签文本分类的复杂计算任务,引入多任务机制,结合标签嵌入表示与文本嵌入表示的相似度计算任务,进一步提高多标签文本分类任务的实验精度。

Multi-label text classification is a challenging task in natural language processing.Traditional text classification tasks can get good predictions simply by associating the features in each sample with a particular label.Each sample in the multi-label text classification task corresponds to one or more labels,and the features in each sample cannot be clearly mapped to the corresponding labels.Aiming at the complex computing task of multi-label text classification,this paper introduces multi-task mechanism that combines the similarity computing task of embedded label representation and embedded text representation to further improve the experimental accuracy of multi-label text classification task.