聊天机器人中用户出行消费意图识别方法

发布时间：2018-02-01 21:18

本文关键词： 出行消费意图识别机器学习深度学习卷积神经网络长短期记忆网络　出处：《中国科学:信息科学》2017年08期 　论文类型：期刊论文

【摘要】：聊天机器人中的出行消费意图是指用户为了满足出行的需要,通过文本表达出对出行类产品或者服务的购买意愿.识别出用户的消费意图可以进行相应的产品推荐,增强用户体验.传统的消费意图识别主要使用基于模板匹配或者基于人工特征集合的机器学习方法,这类方法费时费力,扩展性不强.本文将出行消费意图识别任务看成一个分类问题,结合深度学习方法识别用户的出行消费意图,该方法不需要人工构造特征集合或匹配模板.具体而言,本文构建了基于卷积的长短期记忆神经网络(Convolutional-LSTM)模型进行出行消费意图识别,首先通过卷积神经网络(CNN)对用户的聊天文本进行特征抽取,随后进行特征组合并送入长短记忆神经网络(LSTM)进行特征表示学习,最后输出分类结果.实验结果表明,在出行消费意图识别任务上,基于Convolutional-LSTM的模型在F值上优于最好的基线方法 2个百分点.
[Abstract]:The purpose of travel consumption in chat robot is to meet the needs of travel. Through the text to express the desire to purchase the travel products or services and identify the consumer intention of the user can carry out the corresponding product recommendation. Enhance the user experience. The traditional consumer intention recognition mainly uses the machine learning method based on template matching or human feature set, which is time-consuming and laborious. In this paper, the identification task of travel consumption intention is regarded as a classification problem, combined with the in-depth learning method to identify the user's travel consumption intention. The method does not need to manually construct feature sets or matching templates. In this paper, a convolution-based long-term and short-term memory neural network Convolutional-LSTM model is constructed to identify travel consumption intention. Firstly, the user's chat text is extracted by the convolution neural network (CNN), and then the feature combination is carried out and the feature representation learning is carried out by the long memory neural network (LSTM). Finally, the classification results are outputted. The experimental results show that, in the travel consumption intention recognition task. The Convolutional-LSTM-based model outperforms the best baseline method by 2 percentage points in F value.
【作者单位】：哈尔滨工业大学计算机科学与技术学院信息检索研究中心;
【基金】：国家重点基础研究发展计划(973)(批准号:2014CB340503) 国家自然科学基金(批准号:61472107,61632011)资助项目
【分类号】：TP18;TP391.1
【正文快照】： 1引言随着社会的日益信息化,人们希望能用自然语言与计算机交流,于是聊天机器人应运而生.在和机器人聊天的过程中,用户产生了大量的数据.这些数据包含了用户丰富的信息,展示了用户的需求和意图,识别出用户的出行意图可以进行相应产品的推荐,使得聊天机器人变得更加智能,增强用

【相似文献】