更改

*On '''Multi-label Text Classification (MLTC),''' text features can be regarded as '''detailed description of documents''' and label sets can be '''a summarization of documents'''. '''Hybrid Topics''' from text features and label sets by LDA (a method of '''topic model''') can effectively mine global label correlations and deeper features.  The paper '''"An Efficient Framework for Multi-label Text Classification"'''  is has been accepted in '''IJCNN 2019'''. *Deep learning For multi-label text classification. We utilize '''dilated convolution''' to obtain the '''semantic understanding''' of the text and design a hybrid '''attention mechansim''' for '''different labels''' (Specifically, each label should attend to most relevant textual informationcontents).  FirstFirstly,ly  we initialize trainable label embeddings. Then After obationing word-level information based on Bi-LSTM, we get semantic understanding of texts based on word-level information by dilated convolution. Finally, we design a hybrid attention for different labels based on label embeddings.  Besides, we add '''label cooccurrence matrix into loss function '''to guide the whole network to learn and achieve good results.   The paper "'''Multi-label Classification: Select Distinct Semantic Understanding of Text for Different Labels'''" has been accepted in'''APWEB-WAIM 2019.''' *'''GCN (Graph Convolution Network) '''can be used to exploit more complex label correlations on Image Multi-label Learning.