摘要
为充分了解顾客对餐品的满意程度,帮助商家准确把握顾客的消费需求,以外卖平台用餐评论数据为基础,采用word2vec_BiLSTM文本情感分类模型的方法,使用word2vec预训练出各评论语句表征的词向量,利用三种基线模型RNN、LSTM、BiLSTM进行对比试验,根据相应的评价指标对多种分类模型效果进行分析。试验结果表明,word2vec_BiLSTM的F_(1)指标为91.71%,与RNN和LSTM模型相比,分别提高了3.81%、2.46%,word2vec_BiLSTM的ACC值为91.19%,与RNN和LSTM模型相比,分别提高了4.56%、1.62%。
In order to fully understand the level of customer satisfaction with the meals and help merchants accurately grasp the consumer demand.Adopting word2vec_BiLSTM text sentiment classification model approach based on takeaway platform dining review data.The word vectors for each comment statement representation were pre-trained using word2vec,and the three baseline models RNN,LSTM and BiLSTM were used for comparison experiments to analyse the effectiveness of the various classification models according to the corresponding evaluation metrics.The experimental results showed that word2vec_BiLstm was better than RNN and LSTM models,word2vec_BiLSTM increased the rating index F1 by 3.81%and 2.46%respectively,the F1 value is 91.71%,the evaluation index ACC by 4.56%and 1.62%respectively,and the ACC value is 91.19%.
作者
秦精俏
王彤
王玉珍
QIN Jingqiao;WANG Tong;WANG Yuzhen(Institute of Information Engineering,Lanzhou University of Finance and economics,Lanzhou 730200,China)
出处
《枣庄学院学报》
2022年第2期37-44,共8页
Journal of Zaozhuang University
基金
甘肃省软科学项目(20CX9ZA062)。