Due to the complexity and variety of domain-specific terms in architectural questions,the common text classification algorithms are more difficult to classify architectural questions.In order to improve the classification performance of questions in the architectural field,this paper proposes an architectural text classification algorithm based on the fusion of RoBERTa and Word2Vec.Experimental results show that the accuracy rate of the proposed method reaches 91.59％on the construction do-main problem dataset,and the classification performance is better,and on general data sets,the accuracy rate is higher than that of SVM,CNN and other models.
关键词
文本分类/预训练语言模型/句向量/深度学习/问答系统
Key words
text classification/pretrained language model/sentence vector/deep learning/question-answering system
维普
万方数据