改进的多模型融合技术在客服问答系统上的应用

Application of improved multi-model fusion technology in customer service answering system

随着人工智能技术的发展,越来越多的公司采用机器客服代替人工客服。但若采用传统关键词模型,则机器客服准确率难以提高;若采用深度学习模型进行训练,则又面临用户问题是短文本时,模型训练和预测效果不佳的问题。针对这些问题,通过深入研究和多次试验,提出一种融合关键词模型和基于字向量的深度学习模型的算法。最后实现了模型的训练和预测,在与传统算法的准确率对比方面展现了优势。

With the development of artificial intelligence(AI), more and more companies use machine customer service instead of manual customer service. However, if the traditional keyword model is adopted, the accuracy of the machine customer service is difficult to improve. If the deep learning model is used, the predict result is poor when the user problem is short text. Aiming at these problems, an algorithm combining keyword model and deep learning model based on word vector was proposed. The training and prediction of the model was realized, and the advantages were shown in the comparison with the accuracy of the traditional algorithm.