【摘要】 聊天机器人在学术界及工业界均受到了广泛的关注.目前在学术界,关于端到端对话回复的研究成果众多.其中,采用数据驱动的对话回复研究方法占主要地位,且多基于深度神经网络学习与理解自然语言.已有的对话回复模型多针对开放领域.在聊天机器人中比较成熟的应用目前也多为娱乐型聊天机器人.专业领域内的聊天机器人(如心理咨询聊天机器人)目前还多基于规则及模板.为了提高心理咨询类聊天机器人的智能性,提出一种应用于校园心理咨询场景下的对话匹配度建模方法,该方法基于心理咨询网站及贴吧语料,提取单词及句子在心理咨询类别上的相关特征,并将此特征应用于机器学习及深度学习网络中进行句对匹配度建模.与传统的开放领域内句对匹配模型相比,该模型利用了心理咨询的领域分析知识,能够达到更好的匹配效果.更多还原

【Abstract】 Chat-bots have received wide attention in both academia and industry.In academia,there have been many promising research results in the end-to-end dialogue response area.Among them,data-driven dialogue response methods predominate,which learn and understand natural language through deep neural networks.Existing dialogue response models are mainly designed for open domains.The current mature chat-bot applications are mostly used for entertainment.Methods used on professional chat-bots(like psychological counseling chat-bots) are mainly based on rule and template.To enhance the intelligence of the psychological counseling chat-bot, a new method of modeling dialogue matching pattern in the context of campus counseling is proposed.This method is based on the psychological counseling website and Tieba corpus,from which relevant characteristics of words and sentences in the category of psychological counseling types are extracted,and are applied to machine learning and deep learning networks to model the dialogue matching pattern.Compared with traditional dialogue matching models in open domain,the proposed model achieved better matching results with the use of analyzed psychological counseling information.更多还原