基于层级图标签表示网络的多标签文本分类

多标签文本分类是一项基础而实用的任务，其目的是为文本分配多个可能的标签。近年来，人们提出了许多基于深度学习的标签关联模型，以结合标签的信息来学习文本的语义表示，取得了良好的分类性能。通过改进标签关联的建模和文本语义表示来推进这一研究方向。一方面，构建的层级图标签表示，除了学习每个标签的局部语义外，还进一步研究多个标签共享的全局语义。另一方面，为了捕捉标签和文本内容间的联系并加以利用，使用标签文本注意机制来引导文本特征的学习过程。在三个多标签基准数据集上的实验表明，该模型与其他方法相比具有更好的分类性能。

Multi-label text classification based on hierarchical graph label representation network

Multi -label text classification is a basic and practical task, and its purpose is to allocate multiple possible labels for text. In recent years, people have proposed a lot of deep -learning label association models, which are expressed by learning the semantics of the text based on the information of labels, and have achieved good classification performance. This paper promoted this research direction by improving the modeling and text semantics of the label association. On the one hand, the constructed hierarchical graph representation not only learned the local semantics of each label, but also further studied the global semantics shared by multiple labels. On the other hand, in order to capture the connection between labels and text content, it used the label text attention mechanism to guide the learning process of text characteristics. Experiments on the three multi -label benchmark data sets show that the proposed model has better classification performance compared to other methods.