Multi-label classification for agricultural text based on ALBERT-Seq2Seq model
To solve the problem that the existing static word vector model cannot capture the complete se-mantics of text for multi-label classification,a neural network model ALBERT-Seq2Seq for agricultural text multi-label classification is proposed in this paper by combining ALBERT with sequence-to-sequence mod-el.The model uses ALBERT pre-training language model to dynamically acquire agricultural text semantics information,uses its internal multi-layer bidirectional Transformer architecture to mine the deep features of agricultural text information,and then constructs a multi-label classifier by introducing Seq2Seq model and trains it.The algorithm performance test on AGRI-ML2020 agricultural text multi-label data set shows that the F1 value of this model reaches 89.5％,which can effectively improve the effect of agricultural text multi-label classification.
natural language processingmulti-label classificationsequence to sequence modelagricul-tural textdeep learning
万方数据
维普
