摘要
传统的文本分类算法采用词向量表示文本,忽视了上下文语境中词义的变化.本文通过引入self-attention机制处理词向量,提出一种卷积神经网络模型与关键词提取技术相结合的文本分类模型.该模型对文档进行self-attention操作,以抽取关键信息,构建文档特征图,根据卷积神经网络模型和关键词提取技术实现特征向量的分类.在真实数据集上进行性能分析,并与循环神经网络模型、长短时记忆网络模型进行比较,结果表明该分类模型有效地提高了分类的准确性.
Word vectors was used to represent text in the traditional text categorization algorithm,ignoring the change of meaning in context. With a self-attention mechanism to process word vectors,this paper proposes a text classification model combining convolutional neural network model and keyword extraction technology. The model performs self-attention operation on the document to extract key information,construct the document feature map,and classify the feature vector according to the convolutional neural network model and keyword extraction technology. The performance analysis is carried out on the real data set,and compared with the cyclic neural network model and the long and short time memory network model. The results show that the classification model effectively improves the classification accuracy.
作者
邵清
马慧萍
SHAO Qing;MA Hui-ping(School of Option-Electrical and Computer Enginnering,University of Shanghai for Science and Technology,Shanghai 200093,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2019年第6期1137-1141,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61703278)资助
上海市科委科研计划项目(17511107203)资助
