融合实体信息的循环神经网络文本分类模型

Text Classification Model Based on Recurrent Neural Network with Entity Information

基于机器学习的文本分类方法通常忽略了文本上下文内容的语义特征,基于深度学习的文本分类方法虽考虑了上下文内容的语义特征,但弱化了实体信息在文本分类中的作用,无法丰富文本语义表示,突出文本内容特征.本文融合实体信息,提出一种循环神经网络与实体表示相结合的分类模型.该模型通过循环神经网络和Attention机制对文本进行建模表示,充分挖掘文本上下文内容的语义特征;同时,将实体表示与文本表示做注意力计算,进一步丰富文本语义与内容特征.在20NG、R8、IMDB、AG News四个数据集上进行实验,实验结果显示,该模型与其他主流基线方法相比,在准确率(Accuracy)上取得显著提高.

The text classification method based on machine learning usually ignores the semantic features of text context content.Although the text classification method based on deep learning takes the semantic features of text context content into account,it weakens the role of entity information in text classification and cannot enrich the text semantic representation and highlight the characteristics of text content.This paper proposes a classification model combining recurrent neural network and entity representation by fusing entity information.This model uses a recurrent neural network and attention mechanism to model text representations to mine semantic features of text context content.At the same time,the entity representation and text representation are used to perform attention calculations to further enrich text semantics and content features.Experiments on four data sets,such as 20 New groups,R8,IM DB,and AG New s,the experimental results show that the model has a significant improvement in accuracy compared with other mainstream baseline methods.