摘要
特定目标情感分类不仅依赖于上下文信息,还需结合特定目标的特征信息,是一种细粒度的情感分析。针对特定目标情感分类提出了一种基于深度记忆网络的分类模型。该模型以双向LSTM和注意力机制为主干框架,从双向LSTM中抽取出目标的特征表示,将目标特征信息加入句子表示中,并加入多计算层(Hops)结构,用以挖掘句子和目标更深层次的情感特征信息,每个计算层的结构类似,共享参数。最后在SemEval2014和SemEval2016数据集上进行实验,取得了比其它基准模型更好的效果。
Aspect-based sentiment classification not only depends on context information,but also needs to combine feature informa?tion of specific aspect.It is a fine-grained emotional analysis.This article proposed a deep memory network model for aspect-based sentiment classification.The model is based on bidirectional LSTM and attention mechanism.The feature representation of the aspect is extracted from the bidirectional LSTM and added to the sentence representation.A multi-computing layer(Hops)structure is intro?duced to mine deeper emotional features of sentences and targets.The structure of each computing layer is similar and parameters are shared.Finally,experiments on SimEval2014 and SimEval2016 datasets show that our model achieves better results than other bench?mark models.
作者
张玲
刘臣
ZHANG Ling;LIU Chen(Business School,University of Shanghai for Science and Technology,Shanghai 200093,China)
出处
《软件导刊》
2019年第12期40-43,50,共5页
Software Guide
关键词
特定目标情感分类
双向LSTM网络
注意力机制
多计算层结构
aspect-based sentiment classification
bidirectional LSTM
attention mechanism
multi-computing layer structure
