摘要
针对旅游网站评论文本的长度短、特征稀疏、多歧义等特征,为了获得更精确的文本分类效果,融合知识增强语义表示(ERNIE)预训练模型和TextRCNN模型,即利用ERNIE模型获得词向量表示,再作为TextRCNN模型的输入,进一步提取文本的上下文信息,获得了效果更好的改进的ERNIE-RCNN短文本分类模型.在旅游评论数据集上的实验证明,改进的ERNIE-RCNN模型具有更好的分类结果.
In view of the short length,sparse features and multiple ambiguities of travel website review text,in order to obtain a more accurate text classification effect,the knowledge enhanced semantic representation(ERNIE)pre-training model is integrated with the TextRCNN model,namely,the ERNIE model is used to obtain the word vector representation,which is then used as the input of the TextRCNN model.The context information of the text is further extracted,and an improved ERNIERCNN model of short text classification with better effect is obtained.Experiments on the travel review dataset show that the improved ERNIE-RCNN model has better classification results.
作者
宋文琴
尚庆生
巩晴
SONG Wenqin;SHANG Qingsheng;GONG Qing(College of Information Engineering,Lanzhou University of Finance and Economics,Lanzhou,Gansu 730030,China)
出处
《宜宾学院学报》
2021年第12期53-56,共4页
Journal of Yibin University
基金
甘肃省自然科学基金项目(21JR1RA283)。