集成学习在短文本分类中的应用研究

Application of ensemble learning in short text classification

为了进一步提高基于深度神经网络短文本分类性能,提出将集成学习方法应用于5种不同的神经网络文本分类器,即卷积神经网络、双向长短时记忆网络、卷积循环神经网络、循环卷积神经网络、分层注意力机制神经网络,分别对两种集成学习方法(Bagging,Stacking)进行了测试。实验结果表明:将多个神经网络短文本分类器进行集成的分类性能要优于单一文本分类模型;进一步两两集成的实验验证了单个模型对短文本分类性能的贡献率。

As for improving the classification performance of short text based on the deep neural networks,the application of ensemble learning method on five different neural network text classifiers(convolutional neural networks,bidirectional long short term memory networks,convolutional cyclic neural networks,recurrent convolutional neural networks,hierarchical attentional neural netwoks) is proposed,in which the two ensemble learning methods(Bagging,Stacking) are tested respectively.The experimental results show that the classification performance of the model integrated with multiple neural network short text classifiers is better than that of any single classification model. The pairwise model combination experiments are carried out to verify the contribution of a single model to short text classification.