营销活动问题标签分类语料库的构建与分类研究

判断营销活动投诉工单所属的标签类别,开展营销活动问题标签分类研究具有重要意义,然而目前尚没有相关语料库。基于K-means算法和专业知识确定分类标签,构建营销活动问题标签分类语料库,且每个问题标签的一致性均达到93%以上。这说明该语料库能够为营销活动投诉工单分类研究提供统一资源支撑。此外,在构建的语料库上,采用单一深度学习模型和融合的方法进行营销活动问题标签分类研究。实验结果显示,F1值达到67.70%,说明该分类方法是有效的。

CONSTRUCTION AND CLASSIFICATION OF QUESTION LABEL CORPUS FOR MARKETING ACTIVITY

It is of great significance to identify the category of complaint order in marketing activities and to carry out label classification research on marketing activities. However, there is no available corpus of complaint orders about marketing activity question. We determined classification labels based on K-means algorithm and professional knowledge, and constructed question label corpus for marketing activity. The consistency of each question label reached over 93%. It showed that the corpus could provide a unified resource support for the research on classification of complaint order in marketing activities. Based on the constructed corpus, we used the single deep learning model and fusion method to classify the question label about marketing activity. The experimental result shows that F1 value reaches 67.70%, which shows that the proposed classification method is effective.