CN105206284A - Virtual chatting method and system relieving psychological pressure of adolescents - Google Patents

Abstract

Translated fromChinese

本发明公开一种疏导青少年心理压力的虚拟聊天方法与系统，旨在通过与青少年进行若干次连续的对话应答，提供一些积极正面的信息，引导他们以一种积极乐观的态度进行思考、面对困难，达到缓解青少年心理压力的目的。所述方法，包括：S1、获取用户的第一会话语句，对第一会话语句进行中文分词，并建立第一依存关系树；S2、通过预先建立的压力词库对第一依存关系树的各个节点进行标注，并计算标注后的压力类型节点与情绪词节点之间的路径长度以及压力子类型节点与情绪词节点之间的路径长度得到中间结果；S3、从中间结果中选出目标中间结果，从目标中间结果中提取目的压力感知结果；S4、根据所述目的压力感知结果得到目标应答语句，并反馈给用户。

The invention discloses a virtual chat method and system for relieving the psychological pressure of teenagers, aiming to provide some positive information through several consecutive dialogue responses with teenagers, and guide them to think and face the situation with a positive and optimistic attitude. Difficulties, to achieve the purpose of alleviating the psychological pressure of young people. The method includes: S1. Obtaining the first conversational sentence of the user, performing Chinese word segmentation on the first conversational sentence, and establishing a first dependency tree; S2. Mark the node, and calculate the path length between the marked pressure type node and the emotional word node and the path length between the pressure subtype node and the emotional word node to obtain an intermediate result; S3, select the target intermediate result from the intermediate result , extracting a target pressure perception result from the target intermediate result; S4, obtaining a target response sentence according to the target pressure perception result, and feeding it back to the user.

Description

Translated fromChinese

疏导青少年心理压力的虚拟聊天方法与系统Virtual chat method and system for alleviating adolescents' psychological pressure

技术领域technical field

本发明涉及数据分析技术领域，具体涉及一种疏导青少年心理压力的虚拟聊天方法与系统。The invention relates to the technical field of data analysis, in particular to a virtual chat method and system for alleviating the psychological pressure of teenagers.

背景技术Background technique

随着社会发展的不断加快，当今人们普遍承受着来自各个方面的心理压力。青少年心智仍未完全成熟，对情绪的把握和自我调节能力与成年人相比相对不足，过大的心理压力往往会对青少年的身心造成不同程度的伤害。同时，处于叛逆期的青少年往往不愿意与他人进行交流，致使其压力情绪无法得到及时的疏导，造成更加严重的后果。相关研究表明，我国10％-30％的青少年有着不同程度的心理问题，九成以上的学生存在学习焦虑心理问题，近半数的学生对考试怀有恐惧心理。与此同时，随着WEB2.0的兴起，网络为越来越多的青少年提供了交流和表达情绪的平台。因此，若能设计并实现一个虚拟聊天工具，通过互联网与青少年进行智能的对话聊天，融合其微博内容，自动识别青少年会话语句中体现的压力情绪和压力来源，给出富有人性化的应答语句来疏导他们的心理压力，则对于青少年的健康成长大有裨益，对整个社会的和谐健康的发展有着不容忽视的重要作用。With the continuous acceleration of social development, people today are generally under psychological pressure from various aspects. Adolescents are still immature mentally, and their ability to grasp emotions and self-regulation is relatively insufficient compared with adults. Excessive psychological pressure often causes varying degrees of physical and mental damage to adolescents. At the same time, teenagers in the rebellious period are often unwilling to communicate with others, resulting in the inability to relieve their stressful emotions in a timely manner, resulting in more serious consequences. Relevant studies have shown that 10%-30% of teenagers in my country have different degrees of psychological problems, more than 90% of students have learning anxiety problems, and nearly half of students have fear of exams. At the same time, with the rise of WEB2.0, the Internet provides a platform for more and more young people to communicate and express their emotions. Therefore, if a virtual chat tool can be designed and implemented to conduct intelligent conversations with teenagers through the Internet, integrate their microblog content, automatically identify the stressful emotions and sources of stress reflected in teenagers' conversational sentences, and give humanized response sentences It is of great benefit to the healthy growth of young people, and plays an important role in the harmonious and healthy development of the whole society that cannot be ignored.

自20世纪90年代开始，人们就开始研究和设计自动问答系统。自动问答系统进行一个问答过程往往需要经历问题文本分析、信息检索、结果评估和排名、整合答案等过程。在自然语言问答系统方面,比较著名的系统有Start、AskJeeves、AnswerBus等。Start系统是由麻省理工学院研发的一款问答系统，能够回答历史、地理科技文化等方面的问题，但是在对上下文的理解方面还有所欠缺。AskJeeves系统的主要功能是检索，通过用户交互和语义分析给出检索结果。AnswerBus是密歇根大学研发的多语言自动问答系统，它在分析用户输入的自然语言句后，给出可能包含答案的8条语句和相关联的URL。此外，微软公司的Encarta、华盛顿大学的MULDER、新加坡国立大学的LAM等也都是相对成熟的自动问答系统。Since the 1990s, people have been researching and designing automatic question answering systems. An automatic question answering system often needs to go through the process of question text analysis, information retrieval, result evaluation and ranking, and integration of answers in a question answering process. In terms of natural language question answering systems, the more famous systems include Start, AskJeeves, AnswerBus and so on. The Start system is a question-answering system developed by the Massachusetts Institute of Technology. It can answer questions about history, geography, technology and culture, but it is still lacking in understanding the context. The main function of the AskJeeves system is retrieval, which provides retrieval results through user interaction and semantic analysis. AnswerBus is a multilingual automatic question answering system developed by the University of Michigan. After analyzing the natural language sentences entered by users, it gives 8 sentences and associated URLs that may contain answers. In addition, Microsoft's Encarta, Washington University's MULDER, and the National University of Singapore's LAM are also relatively mature automatic question answering systems.

近年来，国内的许多科研机构和企业也对以汉语为主要语言的问答系统进行了相关研究。2013年，哈尔滨工业大学在ACL上发表的文章《AChatterbotSystemforAnsweringDomain-specificQuestions》介绍了他们设计的围绕情感领域的自动问答系统。该问答系统将用户提出的问题与数据库中的问题进行匹配，筛选出相似度最大的问题并给出相应解答。In recent years, many domestic scientific research institutions and enterprises have also conducted related research on question answering systems with Chinese as the main language. In 2013, the article "AChatterbot System for Answering Domain-specific Questions" published by Harbin Institute of Technology on ACL introduced the automatic question answering system they designed around the emotional field. The question answering system matches the questions raised by users with the questions in the database, screens out the most similar questions and gives corresponding answers.

然而智能问答系统存在三个无法应用于青少年心理压力疏导的问题：1)自动问答系统往往针对于已有客观或公认答案的问题，无法满足用户在有压力情绪时的需求；2)“问答”模式缺少互动：在压力疏导的过程中，“提问者”和“回答者”的角色应该在用户和聊天程序中进行适当转换，以提高整个疏导过程的互动性；3)“问答”模式缺少对语言环境的理解：用户的倾诉、聊天过程是一个连续性的过程，每一句会话语句都有着上下文的语境。若按照自动问答系统的工作原理，只是单纯对每一句语句进行分别应答，则应答效果将深受局限。However, there are three problems in the intelligent question answering system that cannot be applied to the psychological stress relief of adolescents: 1) The automatic question answering system is often aimed at questions that have objective or accepted answers, and cannot meet the needs of users when they are under stress; 2) "Question answering" Mode lacks interaction: In the process of stress relief, the roles of "asker" and "answerer" should be properly switched between users and chat programs to improve the interactivity of the entire counseling process; 3) "Question and answer" mode lacks support for Understanding of the language environment: The user's talk and chat process is a continuous process, and each conversational sentence has a context. If according to the working principle of the automatic question answering system, each sentence is simply answered separately, the answer effect will be deeply limited.

智能聊天机器人是自动问答系统的一个分支，一般基于模式匹配的方来寻找一个恰当的回答，对用户的自然语言输入给出一个相对人性化的回答，这个答案可能并不具有知识性，更多的是模拟人与人聊天交谈的特点。目前比较成熟的智能聊天机器人有国外的Eliza和Alice，国内的“小i”、“小图”和“小黄鸡”等。Eliza是JosephWeizenbaum在1956年推出的聊天机器人，被认定为世界上第一个聊天机器人。它具有一定的心理学知识，能够和用户进行交流，但在上下文之间欠缺良好的连贯性。Alice是一个基于经验的聊天机器人，由RichardS.Wallace博士在1995年开始研发,它最大的特点在于使用了人工智能标记语言AIML作为知识描述语言，对于用户输入的自然语句，在分词后和知识库中的模板进行匹配，获取答案。“小i”是上海赢思软件技术有限公司自主研发的一款中文聊天机器人，经过十年的发展，它已在天气、地图、翻译等多个领域积累了大量的相关知识，且支持语音识别功能，是一款比较成熟的中文聊天机器人。“小黄鸡”起源于人人网，之后迅速火爆，是一款纯粹的聊天机器人。它的智能依赖于庞大的后台知识库，通过教学功能，不断学习用户答案的关键词和答案并存储。Intelligent chatbot is a branch of automatic question answering system. It generally finds an appropriate answer based on pattern matching, and gives a relatively humanized answer to the user's natural language input. This answer may not be knowledgeable. More The most important thing is to simulate the characteristics of chatting with people. At present, relatively mature intelligent chat robots include Eliza and Alice abroad, and domestic "Xiaoi", "Xiaotu" and "Little Yellow Chicken". Eliza is a chat robot launched by Joseph Weizenbaum in 1956 and is recognized as the world's first chat robot. It has some psychological knowledge and can communicate with users, but it lacks good coherence between contexts. Alice is an experience-based chat robot, which was developed by Dr. Richard S. Wallace in 1995. Its biggest feature is that it uses the artificial intelligence markup language AIML as a knowledge description language. For natural sentences entered by users, after word segmentation and knowledge base Match the template in to get the answer. "Xiaoi" is a Chinese chat robot independently developed by Shanghai Yingsi Software Technology Co., Ltd. After ten years of development, it has accumulated a lot of relevant knowledge in weather, maps, translation and other fields, and supports speech recognition It is a relatively mature Chinese chat robot. "Little Yellow Chicken" originated from Renren, and quickly became popular. It is a pure chat robot. Its intelligence relies on a huge background knowledge base. Through the teaching function, it continuously learns and stores the keywords and answers of user answers.

智能聊天机器人在处理用户压力情绪上的主要问题在于：1)面对用户饱含负面情绪的会话语句，多款智能聊天机器人给出的应答不够人性化；2)难以给出切实的解决方案：比如当用户遇到学业压力时，聊天工具应向用户提供一些学习方法上的指导。而在尝试了若干款流行的聊天机器人后，发现当向它咨询“我该怎么办”类似的问题时，其回答往往含糊其辞，无法给出切实的意见。The main problems of intelligent chatbots in dealing with users’ stress are: 1) In the face of users’ conversational sentences full of negative emotions, the answers given by various intelligent chatbots are not humane enough; 2) It is difficult to give practical solutions: for example When users encounter academic pressure, the chat tool should provide users with some guidance on learning methods. After trying several popular chatbots, I found that when I asked it questions like "what should I do", its answers were often vague and unable to give tangible opinions.

有鉴于此，为青少年提供一种以疏导青少年心理压力为目的的智能聊天工具成为亟待解决的问题。In view of this, it is an urgent problem to provide teenagers with an intelligent chat tool for the purpose of relieving their psychological pressure.

发明内容Contents of the invention

本发明的目的在于，提供一种疏导青少年心理压力的虚拟聊天方法与系统，通过与青少年进行若干次连续的对话应答，感知用户的压力情绪和压力来源信息，并根据用户的压力情绪和压力来源信息，及时在线予以沟通、帮助，提供一些积极正面的信息，引导他们以一种积极乐观的态度进行思考、面对困难，达到缓解青少年心理压力的目的。The purpose of the present invention is to provide a virtual chat method and system for relieving the psychological pressure of teenagers. Through several consecutive dialogue responses with teenagers, the user's stress emotion and stress source information can be sensed, and according to the user's stress emotion and stress source Information, online communication and help in a timely manner, provide some positive information, guide them to think with a positive and optimistic attitude, face difficulties, and achieve the purpose of alleviating the psychological pressure of young people.

为此目的，一方面，本发明提出一种疏导青少年心理压力的虚拟聊天方法，包括：For this purpose, on the one hand, the present invention proposes a kind of virtual chat method of dredging young people's psychological pressure, comprising:

S1、获取用户的第一会话语句，对所述第一会话语句进行中文分词，并建立第一依存关系树，其中，所述第一依存关系树的根节点是所述第一会话语句的中心词，各个子节点为所述进行中文分词得到的词，每一条边表示该边相应的子节点的依存关系；S1. Obtain the user's first conversational sentence, perform Chinese word segmentation on the first conversational sentence, and establish a first dependency tree, wherein the root node of the first dependency tree is the center of the first conversational sentence Word, each sub-node is the word obtained by performing Chinese word segmentation, and each edge represents the dependency relationship of the corresponding sub-node of the edge;

S2、通过预先建立的压力词库对所述第一依存关系树的各个节点进行标注，并计算标注后的压力类型节点与情绪词节点之间的路径长度以及压力子类型节点与情绪词节点之间的路径长度得到中间结果，其中，所述压力词库包括否定词以及其依存的正面情绪词、压力情绪词、压力类型词、压力类型词下的压力子类型词和属于压力子类型词所对应的压力子类型范畴的描述词，所述中间结果的格式为(Stress，Category：l_c，Subcategory：l_s)，Stress表示是否有压力，Category表示压力类型词，l_c表示压力类型词与情绪词节点之间的路径长度，Subcategory表示压力子类型词，l_s表示压力子类型词与情绪词节点之间的路径长度；S2. Mark each node of the first dependency tree through the pre-established pressure lexicon, and calculate the path length between the marked pressure type node and the emotional word node and the distance between the pressure subtype node and the emotional word node. The path length between them is used to obtain an intermediate result, wherein the stress lexicon includes negative words and their dependent positive emotional words, stress emotional words, stress type words, stress subtype words under the stress type words, and words belonging to the stress subtype words The descriptor of the corresponding stress subtype category, the format of the intermediate result is (Stress, Category: l_c , Subcategory: l_s ), Stress represents whether there is stress, Category represents the stress type word, l_c represents the pressure type word and The path length between the emotional word nodes, Subcategory represents the pressure subtype word, and l_s represents the path length between the pressure subtype word and the emotional word node;

S3、从所述中间结果中选出目标中间结果，从所述目标中间结果中提取目的压力感知结果，其中，所述目的压力感知结果的格式为(Stress，Category，Subcategory)；S3. Select a target intermediate result from the intermediate results, and extract a target stress perception result from the target intermediate result, wherein the format of the target stress perception result is (Stress, Category, Subcategory);

S4、根据所述目的压力感知结果得到目标应答语句，并反馈给用户。S4. Obtain a target response sentence according to the target pressure perception result, and feed it back to the user.

另一方面，本发明提出一种疏导青少年心理压力的虚拟聊天系统，包括：On the other hand, the present invention proposes a virtual chat system for relieving adolescents' psychological pressure, including:

建立单元，用于获取用户的第一会话语句，对所述第一会话语句进行中文分词，并建立第一依存关系树，其中，所述第一依存关系树的根节点是所述第一会话语句的中心词，各个子节点为所述进行中文分词得到的词，每一条边表示该边相应的子节点的依存关系；The establishment unit is used to obtain the user's first conversation sentence, perform Chinese word segmentation on the first conversation sentence, and establish a first dependency tree, wherein the root node of the first dependency tree is the first conversation The central word of the sentence, each child node is the word obtained by performing Chinese word segmentation, and each edge represents the dependency relationship of the corresponding child nodes of the edge;

计算单元，通过预先建立的压力词库对所述第一依存关系树的各个节点进行标注，并计算标注后的压力类型节点与情绪词节点之间的路径长度以及压力子类型节点与情绪词节点之间的路径长度得到中间结果，其中，所述压力词库包括否定词以及其依存的正面情绪词、压力情绪词、压力类型词、压力类型词下的压力子类型词和属于压力子类型词所对应的压力子类型范畴的描述词，所述中间结果的格式为(Stress，Category：l_c，Subcategory：l_s)，Stress表示是否有压力，Category表示压力类型词，l_c表示压力类型词与情绪词节点之间的路径长度，Subcategory表示压力子类型词，l_s表示压力子类型词与情绪词节点之间的路径长度；The calculation unit is used to mark each node of the first dependency tree through the pre-established pressure lexicon, and calculate the path length between the marked pressure type node and the emotional word node, as well as the pressure subtype node and the emotional word node The path length between obtains the intermediate result, wherein, described stress lexicon comprises negative word and its dependent positive emotion word, stress emotion word, pressure type word, pressure subtype word under pressure type word and belongs to pressure subtype word The descriptor of the corresponding stress subtype category, the format of the intermediate result is (Stress, Category: l_c , Subcategory: l_s ), Stress represents whether there is stress, Category represents the stress type word, and l_c represents the stress type word and the path length between the emotion word node, Subcategory represents the pressure subtype word, and l_s represents the path length between the pressure subtype word and the emotion word node;

提取单元，用于从所述中间结果中选出目标中间结果，从所述目标中间结果中提取目的压力感知结果，其中，所述目的压力感知结果的格式为(Stress，Category，Subcategory)；An extraction unit, configured to select a target intermediate result from the intermediate results, and extract a target stress perception result from the target intermediate result, wherein the format of the target stress perception result is (Stress, Category, Subcategory);

获取单元，用于根据所述目的压力感知结果得到目标应答语句，并反馈给用户。An acquisition unit, configured to obtain a target response sentence according to the target pressure perception result, and give feedback to the user.

本发明实施例所述的疏导青少年心理压力的虚拟聊天方法与系统，通过感知用户每一次会话语句所蕴含的情绪词、压力类型词和压力子类型词及其依存关系得到目的压力感知结果，并根据所述目的压力感知结果得到相应的目标应答语句，并反馈给用户，能够通过与青少年进行若干次连续的对话应答，为其提供一些积极正面的信息，引导他们以一种积极乐观的态度进行思考、面对困难，达到缓解青少年心理压力的目的。The virtual chat method and system for alleviating the psychological pressure of teenagers described in the embodiment of the present invention obtains the target pressure perception result by perceiving the emotional words, pressure type words and pressure subtype words contained in each conversation sentence of the user and their dependencies, and According to the target pressure perception results, the corresponding target response sentence is obtained and fed back to the user. Through several consecutive conversations and responses with the teenagers, some positive information can be provided to guide them to develop in a positive and optimistic manner. Thinking and facing difficulties to achieve the purpose of relieving the psychological pressure of young people.

附图说明Description of drawings

图1为本发明疏导青少年心理压力的虚拟聊天方法一实施例的流程示意图；Fig. 1 is the schematic flow chart of an embodiment of the virtual chat method of the present invention to relieve the psychological stress of teenagers;

图2为本发明疏导青少年心理压力的虚拟聊天方法另一实施例的压力词库示意图；Fig. 2 is the schematic diagram of the stress lexicon of another embodiment of the virtual chat method of the present invention to relieve the psychological pressure of teenagers;

图3为本发明疏导青少年心理压力的虚拟聊天方法又一实施例中第一依存关系树的示意图；Fig. 3 is a schematic diagram of the first dependency tree in another embodiment of the virtual chat method for relieving adolescents' psychological pressure according to the present invention;

图4为本发明疏导青少年心理压力的虚拟聊天方法又一实施例中行人间法向重叠及切向位移示意图；Fig. 4 is a schematic diagram of the normal overlap and tangential displacement between pedestrians in another embodiment of the virtual chat method for relieving adolescents' psychological pressure according to the present invention;

图5为本发明疏导青少年心理压力的虚拟聊天系统一实施例的方框结构示意图；Fig. 5 is the schematic diagram of block structure of an embodiment of the virtual chat system of the present invention to relieve the psychological pressure of teenagers;

图6为本发明疏导青少年心理压力的虚拟聊天系统另一实施例的架构图；Fig. 6 is the frame diagram of another embodiment of the virtual chat system of the present invention to relieve the psychological pressure of teenagers;

图7为本发明疏导青少年心理压力的虚拟聊天系统又一实施例的登陆界面示意图。Fig. 7 is a schematic diagram of the login interface of another embodiment of the virtual chat system for alleviating the psychological stress of teenagers according to the present invention.

具体实施方式Detailed ways

为使本发明实施例的目的、技术方案和优点更加清楚，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚地描述，显然，所描述的实施例是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are the Some, but not all, embodiments are invented. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

如图1所示，本实施例公开一种疏导青少年心理压力的虚拟聊天方法，包括：As shown in Figure 1, the present embodiment discloses a virtual chat method for relieving the psychological pressure of teenagers, including:

S1、获取用户的第一会话语句，对所述第一会话语句进行中文分词，并建立第一依存关系树，其中，所述第一依存关系树的根节点是所述第一会话语句的中心词，各个子节点为所述进行中文分词得到的词，每一条边表示该边相应的子节点的依存关系；S1. Obtain the user's first conversational sentence, perform Chinese word segmentation on the first conversational sentence, and establish a first dependency tree, wherein the root node of the first dependency tree is the center of the first conversational sentence Word, each sub-node is the word obtained by performing Chinese word segmentation, and each edge represents the dependency relationship of the corresponding sub-node of the edge;

S2、通过预先建立的压力词库对所述第一依存关系树的各个节点进行标注，并计算标注后的压力类型节点与情绪词节点之间的路径长度以及压力子类型节点与情绪词节点之间的路径长度得到中间结果，其中，所述压力词库包括否定词以及其依存的正面情绪词、压力情绪词、压力类型词、压力类型词下的压力子类型词和属于压力子类型词所对应的压力子类型范畴的描述词，所述中间结果的格式为(Stress，Category：l_c，Subcategory：l_s)，Stress表示是否有压力，Category表示压力类型词，l_c表示压力类型词与情绪词节点之间的路径长度，Subcategory表示压力子类型词，l_s表示压力子类型词与情绪词节点之间的路径长度；S2. Mark each node of the first dependency tree through the pre-established pressure lexicon, and calculate the path length between the marked pressure type node and the emotional word node and the distance between the pressure subtype node and the emotional word node. The path length between them is used to obtain an intermediate result, wherein the stress lexicon includes negative words and their dependent positive emotional words, stress emotional words, stress type words, stress subtype words under the stress type words, and words belonging to the stress subtype words The descriptor of the corresponding stress subtype category, the format of the intermediate result is (Stress, Category: l_c , Subcategory: l_s ), Stress represents whether there is stress, Category represents the stress type word, l_c represents the pressure type word and The path length between the emotional word nodes, Subcategory represents the pressure subtype word, and l_s represents the path length between the pressure subtype word and the emotional word node;

S3、从所述中间结果中选出目标中间结果，从所述目标中间结果中提取目的压力感知结果，其中，所述目的压力感知结果的格式为(Stress，Category，Subcategory)；S3. Select a target intermediate result from the intermediate results, and extract a target stress perception result from the target intermediate result, wherein the format of the target stress perception result is (Stress, Category, Subcategory);

S4、根据所述目的压力感知结果得到目标应答语句，并反馈给用户。S4. Obtain a target response sentence according to the target pressure perception result, and feed it back to the user.

本发明实施例中，针对用户输入的会话语句c,可以定义目标压力感知结果为(Stress，Category，Subcategory)。为了获得(Stress，Category，Subcategory)形式的目标压力感知结果，可以首先利用语言分析工具LTP对该语句c进行中文分词并基于依存语法理论建立一棵依存关系树。根据依存语法，句子是一个有组织的整体，构成的成分是词，词与词之间会产生联系，这些词间联系构成了句子的框架并表达思想。因此在该树中，每一个节点是进行中文分词后的词，其根节点是整个句子中的中心词，每一条边表示两个词之间的依存关系，词与词之间的依存关系反应了词与词之间的语义关系。基于这棵依存关系树，首先根据已经建立好的压力词库(如图2的表1所示，包括压力情绪词库和压力类型/子类型词库)对于每一个节点进行标注。如果有压力情绪词存在或者否定词和其依存的正面情绪词存在则Stress取值为1，反之则为0。接下来，分析所有压力情绪词和压力类型词之间的连接，并获得了一系列形如(Stress，Category：l_c，Subcategory：l_s)的中间结果。其中Stress，Category，Subcategory仍然分别表示是否有压力，压力类型和子类型，l_c和l_s压力类型节点和压力子类型节点与压力情绪词节点的路径长度(l_c/l_s可以取值为0，表示在压力类型节点/压力子类型节点与情绪词节点之间没有路径)。最后，优先选取l_c值最小的中间结果作为目标中间结果，若l_c值最小的中间结果有多个，则选取l_s值最小的中间结果作为目标中间结果。但在某些情况下，l_c和l_s值都最小的中间结果仍然有多个，这时选择与历史压力类型相同的且具有最小路径的中间结果作为目标中间结果。而当历史信息不能帮助选择时，则随机选择一个具有最小路径的中间结果作为目标中间结果。综上，对于用户的一个输入会话语句c，将获得一个三维的目标压力感知结果(c.Stress，c.Category，c.Subcategory)。In the embodiment of the present invention, for the conversational statement c input by the user, the target stress perception result may be defined as (Stress, Category, Subcategory). In order to obtain the target stress perception result in the form of (Stress, Category, Subcategory), the language analysis tool LTP can be used to segment the sentence c in Chinese and a dependency tree is established based on the dependency grammar theory. According to dependency grammar, a sentence is an organized whole, composed of words, and there will be connections between words, and these connections between words constitute the framework of the sentence and express ideas. Therefore, in this tree, each node is a word after Chinese word segmentation, and its root node is the central word in the entire sentence. Each edge represents the dependency relationship between two words, and the dependency relationship between words reflects Semantic relationship between words. Based on this dependency tree, firstly mark each node according to the established stress lexicon (as shown in Table 1 of FIG. 2 , including the stress emotion lexicon and the stress type/subtype lexicon). If there are stress emotional words or negative words and their dependent positive emotional words, the value of Stress is 1, otherwise it is 0. Next, the connection between all stress emotion words and stress type words is analyzed, and a series of intermediate results such as (Stress, Category: l_c , Subcategory: l_s ) are obtained. Among them, Stress, Category, and Subcategory still respectively indicate whether there is stress, stress type and subtype, l_c and l_s are the path lengths between stress type nodes and stress subtype nodes and stress emotion word nodes (l_c /l_s can take a value of 0 , indicating that there is no path between the stress type node/stress subtype node and the emotion word node). Finally, the intermediate result with the smallest l_c value is selected first as the target intermediate result. If there are multiple intermediate results with the smallest l_c value, the intermediate result with the smallest l_s value is selected as the target intermediate result. However, in some cases, there are still multiple intermediate results with the smallest values of_lc and_ls . At this time, the intermediate result with the same type of historical pressure and the smallest path is selected as the target intermediate result. And when historical information cannot help the selection, an intermediate result with the smallest path is randomly selected as the target intermediate result. To sum up, for an input conversation sentence c of the user, a three-dimensional target stress perception result (c.Stress, c.Category, c.Subcategory) will be obtained.

本发明实施例所述的疏导青少年心理压力的虚拟聊天方法，通过感知用户每一次会话语句所蕴含的情绪词、压力类型词和压力子类型词及其依存关系得到目的压力感知结果，并根据所述目的压力感知结果得到相应的目标应答语句，并反馈给用户，能够通过与青少年进行若干次连续的对话应答，为其提供一些积极正面的信息，引导他们以一种积极乐观的态度进行思考、面对困难，达到缓解青少年心理压力的目的。The virtual chat method for relieving the psychological pressure of teenagers described in the embodiment of the present invention obtains the target pressure perception result by perceiving the emotional words, pressure type words and pressure subtype words and their dependencies contained in each conversation sentence of the user, and according to the The corresponding target response sentences are obtained by describing the target pressure perception results and fed back to the user. Through several consecutive dialogue responses with teenagers, some positive information can be provided to guide them to think in a positive and optimistic manner. Face difficulties and achieve the purpose of alleviating the psychological pressure of young people.

可选地，在本发明疏导青少年心理压力的虚拟聊天方法的另一实施例中，所述通过预先建立的压力词库对所述第一依存关系树的各个节点进行标注，包括：Optionally, in another embodiment of the virtual chat method for relieving adolescents' psychological stress in the present invention, the labeling of each node of the first dependency tree through the pre-established stress lexicon includes:

通过查找所述压力词库判断所述各个节点中是否存在否定词以及其依存的正面情绪词或者压力情绪词，若存在，则确定Stress的值为有压力所对应的值，否则，确定Stress的值为无压力所对应的值；Judging whether there are negative words and its dependent positive emotional words or stress emotional words in each node by searching the stress lexicon, if there is, then determine that the value of Stress has a value corresponding to pressure, otherwise, determine the value of Stress The value is the value corresponding to no pressure;

对于每一个节点，通过查找所述压力词库确定出该节点所属的压力类型词Category和压力子类型词Subcategory；For each node, determine the pressure type word Category and the pressure subtype word Subcategory to which the node belongs by searching the pressure thesaurus;

根据所述Stress的值、所述各个节点所属的压力类型词Category和压力子类型词Subcategory得到(Stress，Category，Subcategory)格式的值，并利用得到的所述(Stress，Category，Subcategory)格式的值对所述各个节点进行标注。Obtain the value of (Stress, Category, Subcategory) format according to the value of described Stress, the pressure type word Category and the pressure subtype word Subcategory of described each node, and utilize the described (Stress, Category, Subcategory) format that obtains Values label the individual nodes.

可选地，在本发明疏导青少年心理压力的虚拟聊天方法的另一实施例中，所述从所述中间结果中选出目标中间结果，包括：Optionally, in another embodiment of the virtual chat method for relieving adolescents' psychological stress in the present invention, the selection of target intermediate results from the intermediate results includes:

从所述中间结果中选出l_c最小的中间结果；Select an intermediate result with the smallest_lc from the intermediate results;

判断所述l_c最小的中间结果的数量是否为1，若为1，则将所述l_c最小的中间结果作为所述目的中间结果，或者judging whether the number of intermediate results with the smallest_lc is 1, and if it is 1, using the intermediate result with the smallest_lc as the target intermediate result, or

若不为1，则从所述l_c最小的中间结果中选出l_c和l_s最小的中间结果，判断所述l_c和l_s最小的中间结果的数量是否为1，若为1，则将所述l_c和l_s最小的中间结果作为所述目的中间结果，或者If it is not 1, select the smallest intermediate result of_lc and_ls from the smallest intermediate result of_{lc, and judge whether the quantity of the smallest intermediate result of lcandls} is 1, if it is 1, Then use the intermediate result with the smallest l_c and l_s as the target intermediate result, or

若所述l_c和l_s最小的中间结果的数量不为1，则判断所述l_c和l_s最小的中间结果中是否存在与历史目的压力感知结果的压力类型词相同的中间结果，若存在，则从所述l_c和l_s最小的中间结果中与历史目的压力感知结果的压力类型词相同的中间结果中选取一个中间结果作为所述目的中间结果，否则，若不存在，则从所述l_c和l_s最小的中间结果中选取一个中间结果作为所述目的中间结果。If the quantity of the minimum intermediate result of_lc and_ls is not 1, then judge whether there is an intermediate result identical to the pressure type word of the pressure perception result of the historical purpose in the intermediate result of the minimum_lc and_ls , if Exist, then select an intermediate result from the intermediate results that are the same as the pressure type word of the historical purpose pressure perception result from the intermediate results with the smallest l_c and l_s as the intermediate result of the purpose, otherwise, if not exist, then select an intermediate result from An intermediate result among the intermediate results with the smallest l_c and l_s is selected as the target intermediate result.

本发明实施例中，以用户输入的会话语句—“开学以后总和朋友吵架，好烦躁！”为例对本发明所述的目的压力感知结果的获取过程进行说明。如图3的依存关系树所示，“烦躁”标注为压力情绪词，“开学”标注为学业压力类型词，“朋友”标注为人际关系中的朋友压力子类型词，“吵架”标注为人际关系压力类型词。由于存在“烦躁”这个压力情绪词，因此Stress取值为1。而分析压力类型和子类型词与压力情绪词之间的关系，可以得到两个中间结果：(1，学业：3，通用general:0)，(1,人际关系：1，朋友：3)。基于最短路径原则，可以选择(1，人际关系，朋友)作为目标压力感知结果。In the embodiment of the present invention, the process of obtaining the objective pressure perception result of the present invention is described by taking the conversational sentence input by the user—"I always quarrel with my friends after school starts, I'm so annoying!" As shown in the dependency tree in Figure 3, "irritability" is marked as a word of stress, "school" is marked as a word of academic pressure, "friend" is marked as a word of friend pressure in interpersonal relationships, and "quarrel" is marked as a word of interpersonal stress. Relationship stress type words. Stress takes a value of 1 because of the stress emotion word "irritability". By analyzing the relationship between stress type and subtype words and stress emotion words, two intermediate results can be obtained: (1, academic: 3, general: 0), (1, interpersonal relationship: 1, friend: 3). Based on the shortest path principle, (1, interpersonal relationship, friends) can be selected as the target pressure perception result.

然而，实际情况中用户的聊天语句都十分简短，由以上方法感知出的三维压力结果中常常会出现c.Stress＝0或c.Category＝null或c.Subcategory＝null的情况，即用户的输入语句并没有具体的描绘压力状况的情绪词或压力类型词。基于聊天过程中压力情绪的持续性，不断记录和更新用户的历史压力状态(h.Stress，h.Category，h.Subcategory)(由以前的输入语句中感知到用户压力状态结果)，并借此对当前感知的压力状态(c.Stress，c.Category，c.Subcategory)中的缺失值进行补充，具体分为如下情况：However, in actual situations, users’ chat sentences are very short, and c.Stress=0 or c.Category=null or c.Subcategory=null often appear in the three-dimensional stress results perceived by the above methods, that is, the user’s input The sentences did not specifically describe emotional words or stress-type words describing stressful situations. Based on the persistence of stress during the chat process, the user's historical stress state (h.Stress, h.Category, h.Subcategory) is continuously recorded and updated (the result of the user's stress state is perceived from the previous input sentence), and thereby Supplement the missing values in the current perceived stress state (c.Stress, c.Category, c.Subcategory), specifically divided into the following cases:

(1)当c.Stress＝0而h.Stress＝0时，如果c.Category＝h.Category，c.Subcategory＝h.Subcategory，则令c.Stress＝1；(1) When c.Stress=0 and h.Stress=0, if c.Category=h.Category, c.Subcategory=h.Subcategory, then let c.Stress=1;

(2)当c.Stress＝1同时h.Stress＝1，如果c.Category＝null，c.Subcategory＝null或c.Category＝h.Category，c.Subcategory＝null，令c.Category＝h.Category，c.Subcategory＝h.Subcategory。(2) When c.Stress=1 and h.Stress=1, if c.Category=null, c.Subcategory=null or c.Category=h.Category, c.Subcategory=null, let c.Category=h. Category, c.Subcategory=h.Subcategory.

将最终的(c.Stress，c.Category，c.Subcategory)记录为新的历史压力状态(h.Stress，h.Category，h.Subcategory)。Record the final (c.Stress, c.Category, c.Subcategory) as a new historical stress state (h.Stress, h.Category, h.Subcategory).

可选地，在本发明疏导青少年心理压力的虚拟聊天方法的另一实施例中，在所述从所述中间结果中选出目标中间结果之后，从所述目标中间结果中提取目的压力感知结果之前，还包括：Optionally, in another embodiment of the virtual chat method for relieving adolescents' psychological stress in the present invention, after the target intermediate result is selected from the intermediate results, the target stress perception result is extracted from the target intermediate result Previously, also included:

判断所述目标中间结果的Stress的值与历史目的压力感知结果的Stress的值是否为有压力所对应的值，若所述历史目的压力感知结果的Stress的值为有压力所对应的值，所述目标中间结果的Stress的值为无压力所对应的值，则判断所述历史目的压力感知结果的Category的值与所述目标中间结果的Category的值以及所述历史目的压力感知结果的Subcategory的值与所述目标中间结果的Subcategory的值是否相等，若相等，则将所述目标中间结果的Stress的值更新为所述历史目的压力感知结果的Stress的值，或者Judging whether the value of Stress of the target intermediate result and the value of Stress of the historical purpose pressure perception result are values corresponding to pressure, if the value of Stress of the historical purpose pressure perception result is a value corresponding to pressure, then The Stress value of the target intermediate result is the value corresponding to no stress, then judge the value of the Category of the historical target stress perception result and the value of the Category of the target intermediate result and the Subcategory of the historical target stress perception result Whether the value is equal to the value of the Subcategory of the target intermediate result, if they are equal, the value of Stress of the target intermediate result is updated to the value of Stress of the historical target pressure perception result, or

若所述历史目的压力感知结果的Stress的值与所述目标中间结果的Stress的值为有压力所对应的值，则判断所述目标中间结果的Category的值与Subcategory的值是否为空值，若为空值，则将所述目标中间结果的Category的值更新为所述历史目的压力感知结果的Category的值，将所述目标中间结果的Subcategory的值更新为所述历史目的压力感知结果的Subcategory的值，否则，若所述目标中间结果的Category的值不为空值，Subcategory的值为空值，则判断所述历史目的压力感知结果的Category的值与所述目标中间结果的Category的值是否相等，若相等，则将所述目标中间结果的Subcategory的值更新为所述历史目的压力感知结果的Subcategory的值。If the value of Stress of the historical target pressure perception result and the value of Stress of the target intermediate result are values corresponding to pressure, it is judged whether the value of Category and Subcategory of the target intermediate result are null values, If it is a null value, update the Category value of the target intermediate result to the Category value of the historical target pressure perception result, and update the Subcategory value of the target intermediate result to the value of the historical target pressure perception result The value of Subcategory, otherwise, if the value of Category of the target intermediate result is not a null value, the value of Subcategory is a null value, then judge the value of the Category of the pressure perception result of the historical purpose and the Category of the target intermediate result Whether the values are equal, and if they are equal, the value of the Subcategory of the target intermediate result is updated to the value of the Subcategory of the historical target pressure perception result.

可选地，在本发明疏导青少年心理压力的虚拟聊天方法的另一实施例中，所述根据所述目的压力感知结果得到目标应答语句，包括：Optionally, in another embodiment of the virtual chat method for relieving adolescents' psychological stress in the present invention, the target response statement obtained according to the target pressure perception results includes:

确定出所述第一会话语句的类型，其中，所述类型包括疑问句和非疑问句；Determine the type of the first conversational sentence, wherein the type includes interrogative sentences and non-interrogative sentences;

判断所述第一会话语句的类型是否为非疑问句以及目的压力感知结果的Stress的值是否为有压力所对应的值，若所述第一会话语句的类型为非疑问句，目的压力感知结果的Stress的值为有压力所对应的值，或者若所述第一会话语句的类型为疑问句，目的压力感知结果的Stress的值为无压力所对应的值，则查找预先建立的第一知识库，得到所述目的压力感知结果的Category的值和Subcategory的值所对应的应答语句，并将查找得到的应答语句作为所述目标应答语句，其中，所述第一知识库中保存有多条由Category的值、Subcategory的值和应答语句组成的记录，或者Determine whether the type of the first conversational sentence is a non-interrogative sentence and whether the value of Stress of the target pressure perception result is a value corresponding to pressure, if the type of the first conversational sentence is a non-interrogative sentence, the Stress of the purpose pressure perception result The value is the value corresponding to stress, or if the type of the first conversational sentence is an interrogative sentence, the value of Stress of the target pressure perception result is the value corresponding to no pressure, then search the pre-established first knowledge base to obtain The response statement corresponding to the value of Category and the value of Subcategory of the target pressure perception result, and the response statement obtained by searching is used as the target response statement, wherein the first knowledge base stores a plurality of value, the value of Subcategory and the record consisting of the answer statement, or

若所述第一会话语句的类型为非疑问句，目的压力感知结果的Stress的值为无压力所对应的值，则从第二知识库中选取所述目标应答语句，其中，所述第二知识库中保存有多条应答语句，或者If the type of the first conversational sentence is a non-interrogative sentence, and the value of Stress of the target pressure perception result is a value corresponding to no pressure, then the target answer sentence is selected from the second knowledge base, wherein the second knowledge There are multiple answer statements stored in the library, or

若所述第一会话语句的类型为疑问句，目的压力感知结果的Stress的值为有压力所对应的值，则从第三知识库中获取所述目标应答语句，其中，所述第三知识库中包括多个问题语句以及相应的答案。If the type of the first conversational sentence is an interrogative sentence, and the value of Stress of the target pressure perception result is a value corresponding to stress, then the target answer sentence is obtained from the third knowledge base, wherein the third knowledge base Include multiple question statements and corresponding answers.

本发明实施例中，基于用户输入语句的类型和目标压力感知结果(c.Stress，c.Category，c.Subcategory)，本聊天工具的应答策略如图4所示。In the embodiment of the present invention, based on the type of sentence input by the user and the target stress perception result (c.Stress, c.Category, c.Subcategory), the response strategy of the chat tool is shown in FIG. 4 .

情况一：有压力感知到(c.Stress＝1)Situation 1: There is a sense of stress (c.Stress=1)

(1)对于陈述句和反问句，本发明会从第一知识库(可以为本地的知识库)中选择一条合适的应答返回给用户。本地知识库中存储着大量的针对不同压力类型和子类型的能够起到安慰，鼓励，或者引导用户的应答语句。(1) For declarative sentences and rhetorical questions, the present invention will select an appropriate answer from the first knowledge base (which may be a local knowledge base) and return it to the user. The local knowledge base stores a large number of response sentences for different stress types and subtypes that can comfort, encourage, or guide users.

(2)对于疑问句，本聊天工具会从第三知识库(中文的最大问答社区)中搜寻一条最合适的应答语句返回给用户，其具体的工作原理如下：(2) For interrogative sentences, this chat tool will search for a most suitable answer sentence from the third knowledge base (the largest question-and-answer community in Chinese) and return it to the user. The specific working principle is as follows:

令Q_u和Q_b分别为用户输入的问题语句和从第三知识库中搜索到的最相关的问题语句。当MatchDegree(Q_u,Q_b)大于一个阈值(即问题语句Q_u和最相关的问题语句Q_b的匹配度)，且该问题的答案被用户采纳或者至少有一个人赞同时，本发明会根据是否被提问者采纳以及赞同数目选取最好的答案返回给用户。否则，本发明返回到本地知识库中选择一个通用的鼓励话语或者一个笑话返回给用户。Let Q_u and Q_b be the question sentence entered by the user and the most relevant question sentence searched from the third knowledge base, respectively. When MatchDegree(Q_u , Q_b ) is greater than a threshold (that is, the matching degree between the question statement Q_u and the most relevant question statement Q_b ), and the answer to the question is adopted by the user or at least one person agrees, the present invention will be based on Whether it is adopted by the questioner and the number of approvals Select the best answer and return it to the user. Otherwise, the present invention returns to the local knowledge base to select a general encouragement or a joke to return to the user.

其中MatchDegree(Q_u,Q_b)基于Q_u和Q_b的依存关系树中的有效节点对进行计算。令有效节点对为n(n₁,n₂)，n₁为依存关系树中的中心节点即根节点，n₂为与n₁节点具有直接依存关系的节点，它可以为一个动词，名词或者是形容词。假设两个有效节点对为u＝(u₁,u₂)和b＝(b₁,b₂)，则可以定义相似度sim(u,b)如下：Among them, MatchDegree(Q_u , Q_b ) is calculated based on valid node pairs in the dependency tree of Q_u and Q_b . Let the effective node pair be n(n₁ ,n₂ ), n₁ is the central node in the dependency tree, i.e. the root node, n₂ is the node that has a direct dependency relationship with n₁ node, it can be a verb, a noun or is an adjective. Assuming that two effective node pairs are u=(u₁ ,u₂ ) and b=(b₁ ,b₂ ), the similarity sim(u,b) can be defined as follows:

$\mathrm{<span><span>s</span>the\; s</span>}\mathrm{<span><span>i</span>i</span>}\mathrm{<span><span>m</span>m</span>}<span><span>(</span>(</span>\mathrm{<span><span>u</span>u</span>}<span><span>,</span>,</span>\mathrm{<span><span>b</span>b</span>}<span><span>)</span>)</span><span><span>=</span>=</span>\mathrm{<span><span>s</span>the\; s</span>}\mathrm{<span><span>i</span>i</span>}\mathrm{<span><span>m</span>m</span>}<span><span>(</span>(</span><span><span>(</span>(</span>{\mathrm{<span><span>u</span>u</span>}}_{\mathrm{<span><span>1</span>1</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>u</span>u</span>}}_{\mathrm{<span><span>2</span>2</span>}}<span><span>)</span>)</span><span><span>,</span>,</span><span><span>(</span>(</span>{\mathrm{<span><span>b</span>b</span>}}_{\mathrm{<span><span>1</span>1</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>b</span>b</span>}}_{\mathrm{<span><span>2</span>2</span>}}<span><span>)</span>)</span><span><span>)</span>)</span><span><span>=</span>=</span>\left\{\begin{array}{cc}\mathrm{<span><span>1</span>1</span>}& \mathrm{<span><span>i</span>i</span>}\mathrm{<span><span>f</span>f</span>}<span><span>(</span>(</span>{\mathrm{<span><span>u</span>u</span>}}_{\mathrm{<span><span>1</span>1</span>}}<span><span>=</span>=</span><span><span>=</span>=</span>{\mathrm{<span><span>b</span>b</span>}}_{\mathrm{<span><span>1</span>1</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>u</span>u</span>}}_{\mathrm{<span><span>2</span>2</span>}}<span><span>=</span>=</span><span><span>=</span>=</span>{\mathrm{<span><span>b</span>b</span>}}_{\mathrm{<span><span>2</span>2</span>}}<span><span>)</span>)</span>\\ \mathrm{<span><span>0.5</span>0.5</span>}& \mathrm{<span><span>i</span>i</span>}\mathrm{<span><span>f</span>f</span>}<span><span>(</span>(</span>{\mathrm{<span><span>u</span>u</span>}}_{\mathrm{<span><span>1</span>1</span>}}<span><span>=</span>=</span><span><span>=</span>=</span>{\mathrm{<span><span>b</span>b</span>}}_{\mathrm{<span><span>1</span>1</span>}}<span><span>,</span>,</span>{\mathrm{<span><span>u</span>u</span>}}_{\mathrm{<span><span>2</span>2</span>}}<span><span>\&NotEqual;</span>\&NotEqual;</span>{\mathrm{<span><span>b</span>b</span>}}_{\mathrm{<span><span>2</span>2</span>}}<span><span>)</span>)</span>\\ \mathrm{<span><span>0</span>0</span>}& \mathrm{<span><span>o</span>o</span>}\mathrm{<span><span>t</span>t</span>}\mathrm{<span><span>h</span>h</span>}\mathrm{<span><span>e</span>e</span>}\mathrm{<span><span>r</span>r</span>}\mathrm{<span><span>w</span>w</span>}\mathrm{<span><span>i</span>i</span>}\mathrm{<span><span>s</span>the\; s</span>}\mathrm{<span><span>e</span>e</span>}\end{array}\right\}<span><span>.</span>.</span>$

令E(Q_u)和E(Q_b)分别为Q_u和Q_b的依存关系树种的有效节点对集。考虑到节点的语义相似性，可以同时用每个节点中词的义原(可以从知网中文词库Hownet中获得)替换掉原始的单词以保证获得更准确的相似度。由此获得有效节点对集为E^*(Q_u)和E^*(Q_b)。由此可以定义MatchDegree(Q_u,Q_b)为：Let E(Q_u ) and E(Q_b ) be the effective node pair sets of the dependency tree species of Qu_u and Q_b respectively. Considering the semantic similarity of the nodes, the original words can be replaced with the sememes of the words in each node (obtained from Hownet Chinese thesaurus Hownet) to ensure a more accurate similarity. Thus, the valid node pair sets are obtained as E^* (Q_u ) and E^* (Q_b ). From this we can define MatchDegree(Q_u ,Q_b ) as:

MatchDegree(Q_u,Q_b)＝αSIM(E(Q_u),E(Q_b))+(1-α)SIM(E^*(Q_u),E^*(Q_b))，$SIM\left(E\right(Q_u),E(Q_b\left)\right)\frac{{\mathrm{\&Sigma;}}_{u\&Element;E\left(Q_u\right)}{\mathrm{\&Sigma;}}_{b\&Element;E\left(Q_b\right)}sim(u,b)}{Max\left\{\right|E\left(Q_u\right)|,|E\left(Q_b\right)\left|\right\}},$$SIM\left(E^{*}\right(Q_u),E^{*}(Q_b\left)\right)=\frac{{\mathrm{\&Sigma;}}_{u^{*}\&Element;E^{*}\left(Q_u\right)}{\mathrm{\&Sigma;}}_{b^{*}\&Element;E^{*}\left(Q_b\right)}sim(u^{*},b^{*})}{Max\left\{\right|E^{*}\left(Q_u\right)\left)\right|,\left|E^{*}\left(Q_b\right)\right|\}}.$MatchDegree(Q_u ,Q_b )＝αSIM(E(Q_u ),E(Q_b ))+(1-α)SIM(E^* (Q_u ),E^* (Q_b )),$SIm\left(\mathrm{E.}\right(Q_u),\mathrm{E.}(Q_b\left)\right)\frac{{\mathrm{\&Sigma;}}_{u\&Element;\mathrm{E.}\left(Q_u\right)}{\mathrm{\&Sigma;}}_{b\&Element;\mathrm{E.}\left(Q_b\right)}\mathrm{the\; s}im(u,b)}{max\left\{\right|\mathrm{E.}\left(Q_u\right)|,|\mathrm{E.}\left(Q_b\right)\left|\right\}},$$SIm\left({\mathrm{E.}}^{*}\right(Q_u),{\mathrm{E.}}^{*}(Q_b\left)\right)=\frac{{\mathrm{\&Sigma;}}_{u^{*}\&Element;{\mathrm{E.}}^{*}\left(Q_u\right)}{\mathrm{\&Sigma;}}_{b^{*}\&Element;{\mathrm{E.}}^{*}\left(Q_b\right)}\mathrm{the\; s}im(u^{*},b^{*})}{max\left\{\right|{\mathrm{E.}}^{*}\left(Q_u\right)\left)\right|,\left|{\mathrm{E.}}^{*}\left(Q_b\right)\right|\}}.$

情况二：没有压力感知到(c.Stress＝0)Case 2: No pressure is sensed (c.Stress=0)

(1)对于非疑问句，本发明将调用开源的聊天机器人simsimi的API，该API接受用户的一条聊天语句输入，返回应答语句；(1) For non-interrogative sentences, the present invention will call the API of the chat robot simsimi of open source, and this API accepts a chat sentence input of the user, returns answer sentence;

(2)对于疑问句，与情况一(2)相同。(2) For interrogative sentences, it is the same as case 1 (2).

可选地，在本发明疏导青少年心理压力的虚拟聊天方法的另一实施例中，所述从第三知识库中获取所述目标应答语句，包括：Optionally, in another embodiment of the virtual chat method for relieving adolescents' psychological stress in the present invention, the acquisition of the target answer sentence from the third knowledge base includes:

从第三知识库中获取与所述第一会话语句最相关的问题语句，其中，所述第三知识库中还包括多个问题语句相应的答案是否被采纳的信息和相应的答案被赞同的数量信息；Obtain the question statement most relevant to the first conversational statement from the third knowledge base, wherein the third knowledge base also includes information on whether the corresponding answers of the plurality of question statements are adopted and whether the corresponding answers are approved quantity information;

计算所述第一会话语句与最相关的问题语句的匹配度，判断所述匹配度是否大于预设的数值，若大于所述数值，则判断所述最相关的问题语句的答案是否被采纳，若被采纳，则从被采纳的答案中选取赞同数量最多的被采纳的答案作为所述目标应答语句，或者，若不大于所述数值或者不被采纳，则从所述第一知识库保存的鼓励话语中随机选取一个鼓励话语作为所述目标应答语句。Calculating the matching degree of the first conversational sentence and the most relevant question sentence, judging whether the matching degree is greater than a preset value, if greater than the numerical value, then judging whether the answer of the most relevant question sentence is adopted, If adopted, select the adopted answer with the largest number of approvals from the adopted answers as the target answer sentence, or, if it is not greater than the numerical value or not adopted, select the answer from the first knowledge base An encouraging utterance is randomly selected from the encouraging utterances as the target response sentence.

可选地，在本发明疏导青少年心理压力的虚拟聊天方法的另一实施例中，所述计算所述第一会话语句与最相关的问题语句的匹配度，包括：Optionally, in another embodiment of the present invention’s virtual chat method for relieving adolescents’ psychological stress, the calculation of the matching degree between the first conversational sentence and the most relevant question sentence includes:

对所述最相关的问题语句Q_b进行中文分词，并建立第二依存关系树；Carry out Chinese word segmentation to described most relevant problem statement Q_b , and set up the second dependency tree;

确定出所述第一依存关系树的有效节点对u＝(u₁,u₂)和第二依存关系树的有效节点对b＝(b₁,b₂)，其中，u＝(u₁,u₂)，u₁为所述第一依存关系树的根节点，u₂为与u₁具有直接依存关系的节点，b₁为所述第二依存关系树的根节点，b₂为与b₁具有直接依存关系的节点；Determine the valid node pair u=(u₁ ,u₂ ) of the first dependency tree and the valid node pair b=(b₁ ,b₂ ) of the second dependency tree, where u=(u₁ , u₂ ), u₁ is the root node of the first dependency tree, u₂ is a node with direct dependency with u₁ , b₁ is the root node of the second dependency tree, b₂ is the node with b₁ Nodes with direct dependencies;

计算所述第一会话语句Q_u与最相关的问题语句Q_b的匹配度，Calculate the degree of matching between the first conversational statement Q_u and the most relevant question statement Q_b ,

计算公式为MatchDegree(Q_u,Q_b)＝αSIM(E(Q_u),E(Q_b))+(1-α)SIM(E^*(Q_u),E^*(Q_b))，其中，MatchDegree(Q_u,Q_b)为所述第一会话语句Q_u与最相关的问题语句Q_b的匹配度，α(α∈(0,1))为常数，E(Q_u)为所述第一依存关系树的有效节点对的集合，E(Q_b)为所述第二依存关系树的有效节点对的集合，E^*(Q_u)为将所述第一依存关系树的各个节点用其义原替代后得到的依存关系树的有效节点对的集合，E^*(Q_b)为将所述第二依存关系树的各个节点用其义原替代后得到的依存关系树的有效节点对的集合，$SIM\left(E\right(Q_u),E(Q_b\left)\right)\frac{{\mathrm{\&Sigma;}}_{u\&Element;E\left(Q_u\right)}{\mathrm{\&Sigma;}}_{b\&Element;E\left(Q_b\right)}sim(u,b)}{Max\left\{\right|E\left(Q_u\right)|,|E\left(Q_b\right)\left|\right\}},$$SIM\left(E^{*}\right(Q_u),E^{*}(Q_b\left)\right)=\frac{{\mathrm{\&Sigma;}}_{u^{*}\&Element;E^{*}\left(Q_u\right)}{\mathrm{\&Sigma;}}_{b^{*}\&Element;E^{*}\left(Q_b\right)}sim(u^{*},b^{*})}{Max\left\{\right|E^{*}\left(Q_u\right)\left)\right|,\left|E^{*}\left(Q_b\right)\right|\}},$u^*为E^*(Q_u)的元素，b^*为E^*(Q_b)的元素，$sim(u,b)=sim\left(\right(u_1,u_2),(b_1,b_2\left)\right)=\left\{\begin{array}{cc}1& if(u_1==b_1,u_2==b_2)\\ 0.5& if(u_1==b_1,u_2\&NotEqual;b_2)\\ 0& otherwise\end{array}\right\}.$The calculation formula is MatchDegree(Q_u ,Q_b )＝αSIM(E(Q_u ),E(Q_b ))+(1-α)SIM(E^* (Q_u ),E^* (Q_b )), where , MatchDegree(Q_u ,Q_b ) is the matching degree between the first conversational sentence Q_u and the most relevant question sentence Q_b , α(α∈(0,1)) is a constant, E(Q_u ) is all The set of valid node pairs of the first dependency tree, E(Q_b ) is the set of valid node pairs of the second dependency tree, E^* (Q_u ) is the set of each of the first dependency tree A collection of effective node pairs of the dependency tree obtained after the node is replaced with its sememe, E^* (Q_b ) is the effective node pair of the dependency tree obtained after each node of the second dependency tree is replaced with its sememe collection of node pairs,$SIm\left(\mathrm{E.}\right(Q_u),\mathrm{E.}(Q_b\left)\right)\frac{{\mathrm{\&Sigma;}}_{u\&Element;\mathrm{E.}\left(Q_u\right)}{\mathrm{\&Sigma;}}_{b\&Element;\mathrm{E.}\left(Q_b\right)}\mathrm{the\; s}im(u,b)}{max\left\{\right|\mathrm{E.}\left(Q_u\right)|,|\mathrm{E.}\left(Q_b\right)\left|\right\}},$$SIm\left({\mathrm{E.}}^{*}\right(Q_u),{\mathrm{E.}}^{*}(Q_b\left)\right)=\frac{{\mathrm{\&Sigma;}}_{u^{*}\&Element;{\mathrm{E.}}^{*}\left(Q_u\right)}{\mathrm{\&Sigma;}}_{b^{*}\&Element;{\mathrm{E.}}^{*}\left(Q_b\right)}\mathrm{the\; s}im(u^{*},b^{*})}{max\left\{\right|{\mathrm{E.}}^{*}\left(Q_u\right)\left)\right|,\left|{\mathrm{E.}}^{*}\left(Q_b\right)\right|\}},$ u^* is an element of E^* (Q_u ), b^* is an element of E^* (Q_b ),$\mathrm{the\; s}im(u,b)=\mathrm{the\; s}im\left(\right(u_1,u_2),(b_1,b_2\left)\right)=\left\{\begin{array}{cc}1& if(u_1==b_1,u_2==b_2)\\ 0.5& if(u_1==b_1,u_2\&NotEqual;b_2)\\ 0& otherwi\mathrm{the\; s}e\end{array}\right\}.$

可选地，在本发明疏导青少年心理压力的虚拟聊天方法的另一实施例中，在所述获取用户的第一会话语句之前，还包括：Optionally, in another embodiment of the virtual chat method for relieving adolescents' psychological stress in the present invention, before acquiring the user's first conversational sentence, it also includes:

接收用户的直接登录指令，并将第一应答语句反馈给用户，或者Receive the user's direct login instruction and feed back the first response statement to the user, or

接收用户的登录指令，并在用户授权连接微博时，获取用户的原创微博，根据所述原创微博判断用户是否有压力情绪，若有，则将第二应答语句反馈给用户，否则，则将第三应答语句反馈给用户。Receive the user's login instruction, and when the user authorizes to connect to the microblog, obtain the user's original microblog, judge whether the user is stressed according to the original microblog, and if so, feed back the second answer statement to the user, otherwise, then Feedback the third answer sentence to the user.

可选地，在本发明疏导青少年心理压力的虚拟聊天方法的另一实施例中，在所述获取用户的第一会话语句之前，还包括：Optionally, in another embodiment of the virtual chat method for relieving adolescents' psychological stress in the present invention, before acquiring the user's first conversational sentence, it also includes:

在将所述第一应答语句、第二应答语句或者第三应答语句反馈给用户之后，获取用户的第二会话语句，对所述第二会话语句进行压力情绪和压力类型检测；After feeding back the first answer sentence, the second answer sentence or the third answer sentence to the user, obtain the user's second conversation sentence, and perform stress emotion and stress type detection on the second conversation sentence;

若所述第二会话语句中检测出压力情绪，未检测出压力类型，则对所述原创微博进行压力检测，并基于所述压力检测结果将第四应答语句反馈给用户，或者If the stress emotion is detected in the second conversation sentence, but the pressure type is not detected, the stress detection is performed on the original microblog, and the fourth response sentence is fed back to the user based on the stress detection result, or

若所述第二会话语句中检测出压力情绪和压力类型或者未检测出压力情绪，则执行步骤S1。If the stress emotion and stress type are detected in the second conversation sentence or the stress emotion is not detected, step S1 is executed.

此外，文本情感分析是对主观性文本进行分析、归纳，获取其情绪特征的过程。通过抽取文本中的情感信息、对情感信息进行分类、分析情感词汇的描述对象来判断短语、句子甚至是整篇文章中的情感倾向。依存文法是由法国语言学L.Tesniere在其编写的《结构句法基础》中提出的。基于微博平台的情感分析工作较为成熟，现有文献(QiLi,YuanyuanXue,JiaJia,LingFeng:HelpingTeenagersRelievePsychologicalPressures:AMicro-blogBasedSystem.[C]EDBT2014:660-663；W.Che,Z.Li,Y.Li,Y.Guo,B.Qin,andT.Liu,“Multilingualdependency-basedsyntacticandsemanticparsing,”inProc.ofCoNLL,2009,pp.49–54.；[15]W.Che,Z.Li,andT.Liu,“LTP:achineselanguagetechnologyplatform,”inProc.ofColing,2010,pp.13–16.)针对青少年所发布的微博，对其中的典型心理压力进行了语义分析，基于微博特征提出青少年心理压力分析的方法，并通过PSS-14心理调查问卷对青少年心理问题进行了较为深入的调研，提出了基于微博“”方式的心理压力疏导工具。本发明对所述原创微博进行压力检测，并基于所述压力检测结果将第四应答语句反馈给用户的实施方式具体可以采用上述现有技术。In addition, text sentiment analysis is the process of analyzing and summarizing subjective texts to obtain their emotional characteristics. By extracting the emotional information in the text, classifying the emotional information, and analyzing the description objects of the emotional vocabulary, we can judge the emotional tendency in phrases, sentences, and even the entire article. Dependency grammar was proposed by French linguistics L. Tesniere in his "Basis of Structural Syntax". Sentiment analysis based on the Weibo platform is relatively mature, and the existing literature (QiLi, YuanyuanXue, JiaJia, LingFeng: Helping Teenagers Relieve Psychological Pressures: AMicro-blogBasedSystem. [C] EDBT2014: 660-663; W.Che, Z.Li, Y.Li, Y. Guo, B. Qin, and T. Liu, “Multilingual dependency-based syntactic and semantic parsing,” in Proc. of CoNLL, 2009, pp.49–54.; [15] W. Che, Z. Li, and T. Liu, “LTP: achinese language technology platform , "inProc.ofColing, 2010, pp.13–16.) Aiming at the microblogs released by teenagers, the typical psychological stress in them was semantically analyzed, and based on the characteristics of microblogs, a method for the analysis of adolescent psychological stress was proposed, and the PSS- The 14 Psychological Survey Questionnaire conducted a relatively in-depth investigation on the psychological problems of adolescents, and proposed a psychological stress relief tool based on the Weibo "" method. The embodiment of the present invention performs pressure detection on the original microblog, and feeds back the fourth response sentence to the user based on the pressure detection result, specifically, the above-mentioned prior art may be adopted.

如图5所示，本实施例公开一种疏导青少年心理压力的虚拟聊天系统，包括：As shown in Figure 5, the present embodiment discloses a virtual chat system for relieving adolescents' psychological pressure, including:

建立单元1，用于获取用户的第一会话语句，对所述第一会话语句进行中文分词，并建立第一依存关系树，其中，所述第一依存关系树的根节点是所述第一会话语句的中心词，各个子节点为所述进行中文分词得到的词，每一条边表示该边相应的子节点的依存关系；The establishment unit 1 is configured to acquire the user's first conversational sentence, perform Chinese word segmentation on the first conversational sentence, and establish a first dependency tree, wherein the root node of the first dependency tree is the first The central word of the conversation sentence, each sub-node is the word that carries out Chinese participle and obtains, and each edge represents the dependency relationship of the corresponding sub-node of this edge;

计算单元2，通过预先建立的压力词库对所述第一依存关系树的各个节点进行标注，并计算标注后的压力类型节点与情绪词节点之间的路径长度以及压力子类型节点与情绪词节点之间的路径长度得到中间结果，其中，所述压力词库包括否定词以及其依存的正面情绪词、压力情绪词、压力类型词、压力类型词下的压力子类型词和属于压力子类型词所对应的压力子类型范畴的描述词，所述中间结果的格式为(Stress，Category：l_c，Subcategory：l_s)，Stress表示是否有压力，Category表示压力类型词，l_c表示压力类型词与情绪词节点之间的路径长度，Subcategory表示压力子类型词，l_s表示压力子类型词与情绪词节点之间的路径长度；Calculation unit 2 is used to mark each node of the first dependency tree through the pre-established pressure lexicon, and calculate the path length between the marked pressure type node and the emotional word node, and the pressure subtype node and the emotional word node. The path length between nodes obtains an intermediate result, wherein the stress lexicon includes negative words and their dependent positive emotional words, stress emotional words, stress type words, stress subtype words under stress type words, and stress subtype words The descriptor of the stress subtype category corresponding to the word, the format of the intermediate result is (Stress, Category: l_c , Subcategory: l_s ), Stress represents whether there is stress, Category represents the stress type word, and l_c represents the stress type The path length between word and emotion word node, Subcategory represents pressure subtype word, l_s represents the path length between pressure subtype word and emotion word node;

提取单元3，用于从所述中间结果中选出目标中间结果，从所述目标中间结果中提取目的压力感知结果，其中，所述目的压力感知结果的格式为(Stress，Category，Subcategory)；An extraction unit 3, configured to select a target intermediate result from the intermediate results, and extract a target stress perception result from the target intermediate result, wherein the format of the target stress perception result is (Stress, Category, Subcategory);

获取单元4，用于根据所述目的压力感知结果得到目标应答语句，并反馈给用户。The obtaining unit 4 is configured to obtain a target response sentence according to the target pressure perception result, and give feedback to the user.

本发明实施例所述的疏导青少年心理压力的虚拟聊天系统，通过感知用户每一次会话语句所蕴含的情绪词、压力类型词和压力子类型词及其依存关系得到目的压力感知结果，并根据所述目的压力感知结果得到相应的目标应答语句，并反馈给用户，能够通过与青少年进行若干次连续的对话应答，为其提供一些积极正面的信息，引导他们以一种积极乐观的态度进行思考、面对困难，达到缓解青少年心理压力的目的。The virtual chat system for alleviating the psychological pressure of teenagers described in the embodiment of the present invention obtains the target pressure perception result by perceiving the emotional words, pressure type words and pressure subtype words and their dependencies contained in each conversation sentence of the user, and according to the The corresponding target response sentences are obtained by describing the target pressure perception results and fed back to the user. Through several consecutive dialogue responses with teenagers, some positive information can be provided to guide them to think in a positive and optimistic manner. Face difficulties and achieve the purpose of alleviating the psychological pressure of young people.

本发明构建了一个用于检测和缓解青少年心理压力的虚拟聊天的解决方案，如图6所示。此系统具体包括如下三大模块：聊天管理器、压力感知器和应答产生器。The present invention constructs a virtual chat solution for detecting and alleviating psychological stress of teenagers, as shown in FIG. 6 . This system specifically includes the following three modules: chat manager, pressure sensor and response generator.

对于聊天管理器，用户登陆系统时，系统将提示用户连接微博(登录界面如图7所示)。若用户没有授权微博链接，则直接进入聊天界面开始常规的聊天程序；当用户选择授权系统微博连接后，网页跳转至聊天页。同时，系统将访问微博相关接口，获取该用户发布的微博。在获得了用户的原创微博后，通过分析用户最近发表的原创微博是否有压力情绪，调整发送给用户的开场白语句。如果系统未授权连接用户微博，或者没有从用户微博中检测到压力情绪，系统将以朋友式的问候用户：“你好！”并开始常规的聊天程序。反之，如果检测到压力情绪，系统将问候用户：“你好！近来过得怎么样？”或“一切安好？”。若用户回复的聊天语句检测出压力情绪但没有压力类型，系统则基于微博的压力检测结果对用户进行尝试性的询问，以引导用户更具体的描述他的压力状况，之后就进入到常规的聊天程序；若用户回复的聊天语句同时检测出压力情绪和压力类型，或者并没有检测出压力情绪，则系统则按常规的聊天程序对用户当前的输入语句进行应答。For the chat manager, when the user logs in to the system, the system will prompt the user to connect to Weibo (the login interface is shown in Figure 7). If the user does not authorize the Weibo link, the user will directly enter the chat interface to start the regular chat program; when the user chooses to authorize the system Weibo link, the web page will jump to the chat page. At the same time, the system will access the Weibo related interface to obtain the Weibo published by the user. After obtaining the user's original microblog, adjust the opening statement sent to the user by analyzing whether the user's recently published original microblog is stressful. If the system is not authorized to connect to the user's Weibo, or if no stress is detected from the user's Weibo, the system will greet the user in a friendly manner: "Hello!" and start a regular chat procedure. Conversely, if a stressful emotion is detected, the system will greet the user: "Hi! How are you doing?" or "Everything is okay?". If the chat sentence replied by the user detects stress but no stress type, the system will make tentative inquiries to the user based on the stress detection results of Weibo to guide the user to describe his stress situation in more detail, and then enter the routine Chat program; if the chat sentence replied by the user detects both stress emotion and stress type, or does not detect stress emotion, then the system responds to the user's current input sentence according to the conventional chat program.

对于压力感知器，为了更全面的理解用户表达的情绪和期望的回答，可以具体设计句子类型识别模块和句子内容分析模块来对用户的聊天语句进行分析。其中句子类型识别模块可以将用户的聊天语句划分为疑问句，反问句和陈述句。例如，“怎么样才能提高我的数学成绩呢？”或者“我应该怎么办呢？”等是疑问句，该用户可能期望获得一些客观的建议或鼓励的话语。而又如“难道他们不应该理解我么？”的反问句则表达了用户一种较强的情绪，他也许是期望获得安抚的回答。在此基础上，句子内容分析模块进一步分析用户输入语句的内容从而来感知用户具体的压力情况(包括压力的类型和子类型)。考虑到用户聊天过程的连续性，该模块同时记录用户历史的压力状态，并借助历史信息来获得更加准确的当前压力感知结果。For the pressure sensor, in order to more comprehensively understand the emotion expressed by the user and the expected answer, the sentence type recognition module and the sentence content analysis module can be specifically designed to analyze the user's chat sentences. The sentence type recognition module can divide the user's chat sentences into interrogative sentences, rhetorical questions and declarative sentences. For example, "How can I improve my math performance?" or "What should I do?" are interrogative sentences, and the user may expect to obtain some objective suggestions or words of encouragement. And rhetorical questions such as "Shouldn't they understand me?" express a strong emotion of the user, and he may expect a soothing answer. On this basis, the sentence content analysis module further analyzes the content of the sentence input by the user to perceive the user's specific stress situation (including the type and subtype of stress). Considering the continuity of the user's chat process, the module also records the user's historical stress state, and uses historical information to obtain more accurate current stress perception results.

对于应答产生器，基于压力感知器分析产生的语句类型结果和压力状态结果，应答选择模块将从本地知识库、第三知识库、以及开源的聊天机器人simsimi库中选择一个合适的回答返回给用户。该模块通过与用户的交互让用户抒发内心的痛苦或者转移情感的注意力从而帮助用户疏导压力。应答管理模块主要负责创建和管理应答策略以及本地知识库。该模块从丰富的网络资源中收集了大量的具有压力疏导功能的应答语句并存储于本地知识库，从而为应答选择模块提供了丰富的应答资源。For the answer generator, based on the sentence type results and stress state results generated by the pressure sensor analysis, the answer selection module will select an appropriate answer from the local knowledge base, the third knowledge base, and the open source chat robot simsimi library to return to the user . This module allows users to express their inner pain or divert their emotional attention through interaction with users to help users relieve stress. The answer management module is mainly responsible for creating and managing answer strategies and local knowledge base. This module collects a large number of answer sentences with pressure relief function from rich network resources and stores them in the local knowledge base, thus providing rich answer resources for the answer selection module.

针对本发明，进行了如下的用户实验：For the present invention, the following user experiments have been carried out:

从中学和大学中邀请了10个年龄介于15岁到22岁之间的学生参与到为期一个月的用户实验中。选择这10个人是因为青少年的主体都是学生而且参与实验的10个人也正遭受着或多或少的来自各个方面的压力。实验一开始，告知实验参与者本聊天工具可以帮助他们疏导心理压力，他们可以在任何时间与我们的聊天工具聊天。测试过程中，实验参与者做以下三件事情：(1)每隔一个星期填写一次压力知觉量表(CPSS-14)(压力知觉量表是心理学上广泛应用于测量人们压力级别的量表)记录他们压力变化情况。。(2)每完成一次会话过程后，针对压力疏导效果给予一个评分。(3)判断每一句的聊天语句的压力感知结果是否正确。Ten middle school and university students aged between 15 and 22 were invited to participate in a one-month user experiment. The reason for choosing these 10 people is that the majority of teenagers are students and the 10 people participating in the experiment are also suffering from more or less pressure from various aspects. At the beginning of the experiment, inform the participants that this chat tool can help them relieve their psychological stress, and they can chat with our chat tool at any time. During the test, the experimental participants did the following three things: (1) Fill out the Stress Perception Scale (CPSS-14) every other week (the Stress Perception Scale is a scale widely used in psychology to measure people's stress levels ) to record their pressure changes. . (2) After each session is completed, a score is given for the effect of stress relief. (3) Judging whether the pressure perception result of each chat sentence is correct.

对于每次会话，将用户的每一条输入语句及其对应的压力感知结果返回给用户，由用户判断压力感知结果是否正确。使用准确率和召回率来评价压力感知算法的精确度。对于压力存在性的感知的查准率和查全率达到78.34％和76.12％。对于压力类型的感知，情感和人际交往类型的召回率以及其他类型的准确率较低，这是因为很多情感和人际交往类型的压力被感知为其他压力类型。另外，对于青少年中最普遍存在的学业压力，本聊天工具具有较好的感知精确度，其查准率和查全率分别达到了86.2％和75％。For each session, each input sentence of the user and its corresponding pressure perception result are returned to the user, and the user judges whether the pressure perception result is correct. Use precision and recall to evaluate the precision of the pressure-aware algorithm. The precision rate and recall rate for the perception of the existence of pressure are 78.34% and 76.12%. For the perception of stress types, recall for the affective and interpersonal types and precision for the other types were lower because many emotional and interpersonal types of stress were perceived as other stress types. In addition, for the most common academic pressure among teenagers, this chat tool has a good perception accuracy, and its precision and recall rates reach 86.2% and 75% respectively.

针对该聊天工具压力疏导的有效性，用户在每一次会话后将针对压力疏导效果给予一个分数(1分：心情更差，2分：心情没有变化，3分：心情感到一点好转，4分：心情感到好转，5分：心情有较大的好转)。有60％以上的会话的得分都超过了2分，即60％以上的会话都能让用户心情好转。本聊天工具对学习类型的压力疏导效果较好，这是因为学习类型的压力感知结果更为准确。Aiming at the effectiveness of the chat tool stress relief, the user will give a score for the effect of stress relief after each session (1 point: the mood is worse, 2 points: the mood has not changed, 3 points: the mood feels a little better, 4 points: The mood feels better, 5 points: the mood has improved a lot). The scores of more than 60% of the sessions are more than 2 points, that is, more than 60% of the sessions can make the user feel better. This chat tool has a better effect on the stress relief of the learning type, because the pressure perception result of the learning type is more accurate.

虽然结合附图描述了本发明的实施方式，但是本领域技术人员可以在不脱离本发明的精神和范围的情况下做出各种修改和变型，这样的修改和变型均落入由所附权利要求所限定的范围之内。Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention. within the bounds of the requirements.

Claims (10)

Translated fromChinese

1.一种疏导青少年心理压力的虚拟聊天方法，其特征在于，包括：1. A virtual chat method for relieving adolescents' psychological pressure, characterized in that, comprising:S1、获取用户的第一会话语句，对所述第一会话语句进行中文分词，并建立第一依存关系树，其中，所述第一依存关系树的根节点是所述第一会话语句的中心词，各个子节点为所述进行中文分词得到的词，每一条边表示该边相应的子节点的依存关系；S1. Obtain the user's first conversational sentence, perform Chinese word segmentation on the first conversational sentence, and establish a first dependency tree, wherein the root node of the first dependency tree is the center of the first conversational sentence Word, each sub-node is the word obtained by performing Chinese word segmentation, and each edge represents the dependency relationship of the corresponding sub-node of the edge;S2、通过预先建立的压力词库对所述第一依存关系树的各个节点进行标注，并计算标注后的压力类型节点与情绪词节点之间的路径长度以及压力子类型节点与情绪词节点之间的路径长度得到中间结果，其中，所述压力词库包括否定词以及其依存的正面情绪词、压力情绪词、压力类型词、压力类型词下的压力子类型词和属于压力子类型词所对应的压力子类型范畴的描述词，所述中间结果的格式为(Stress，Category：l_c，Subcategory：l_s)，Stress表示是否有压力，Category表示压力类型词，l_c表示压力类型词与情绪词节点之间的路径长度，Subcategory表示压力子类型词，l_s表示压力子类型词与情绪词节点之间的路径长度；S2. Mark each node of the first dependency tree through the pre-established pressure lexicon, and calculate the path length between the marked pressure type node and the emotional word node and the distance between the pressure subtype node and the emotional word node. The path length between them is used to obtain an intermediate result, wherein the stress lexicon includes negative words and their dependent positive emotional words, stress emotional words, stress type words, stress subtype words under the stress type words, and words belonging to the stress subtype words The descriptor of the corresponding stress subtype category, the format of the intermediate result is (Stress, Category: l_c , Subcategory: l_s ), Stress represents whether there is stress, Category represents the stress type word, l_c represents the pressure type word and The path length between the emotional word nodes, Subcategory represents the pressure subtype word, and l_s represents the path length between the pressure subtype word and the emotional word node;S3、从所述中间结果中选出目标中间结果，从所述目标中间结果中提取目的压力感知结果，其中，所述目的压力感知结果的格式为(Stress，Category，Subcategory)；S3. Select a target intermediate result from the intermediate results, and extract a target stress perception result from the target intermediate result, wherein the format of the target stress perception result is (Stress, Category, Subcategory);S4、根据所述目的压力感知结果得到目标应答语句，并反馈给用户。S4. Obtain a target response sentence according to the target pressure perception result, and feed it back to the user.2.根据权利要求1所述的疏导青少年心理压力的虚拟聊天方法，其特征在于，所述通过预先建立的压力词库对所述第一依存关系树的各个节点进行标注，包括：2. the virtual chat method of dredging young people's psychological pressure according to claim 1, it is characterized in that, described each node of described first dependency tree is marked by pre-established stress lexicon, comprising:通过查找所述压力词库判断所述各个节点中是否存在否定词以及其依存的正面情绪词或者压力情绪词，若存在，则确定Stress的值为有压力所对应的值，否则，确定Stress的值为无压力所对应的值；Judging whether there are negative words and its dependent positive emotional words or stress emotional words in each node by searching the stress lexicon, if there is, then determine that the value of Stress has a value corresponding to pressure, otherwise, determine the value of Stress The value is the value corresponding to no pressure;对于每一个节点，通过查找所述压力词库确定出该节点所属的压力类型词Category和压力子类型词Subcategory；For each node, determine the pressure type word Category and the pressure subtype word Subcategory to which the node belongs by searching the pressure thesaurus;根据所述Stress的值、所述各个节点所属的压力类型词Category和压力子类型词Subcategory得到(Stress，Category，Subcategory)格式的值，并利用得到的所述(Stress，Category，Subcategory)格式的值对所述各个节点进行标注。Obtain the value of (Stress, Category, Subcategory) format according to the value of described Stress, the pressure type word Category and the pressure subtype word Subcategory of described each node, and utilize the described (Stress, Category, Subcategory) format that obtains Values label the individual nodes.3.根据权利要求1所述的疏导青少年心理压力的虚拟聊天方法，其特征在于，所述从所述中间结果中选出目标中间结果，包括：3. The virtual chat method for dredging young people's psychological pressure according to claim 1, characterized in that, selecting the target intermediate result from the intermediate result includes:从所述中间结果中选出l_c最小的中间结果；Select an intermediate result with the smallest_lc from the intermediate results;判断所述l_c最小的中间结果的数量是否为1，若为1，则将所述l_c最小的中间结果作为所述目的中间结果，或者judging whether the number of intermediate results with the smallest_lc is 1, and if it is 1, using the intermediate result with the smallest_lc as the target intermediate result, or若不为1，则从所述l_c最小的中间结果中选出l_c和l_s最小的中间结果，判断所述l_c和l_s最小的中间结果的数量是否为1，若为1，则将所述l_c和l_s最小的中间结果作为所述目的中间结果，或者If it is not 1, select the smallest intermediate result of_lc and_ls from the smallest intermediate result of_{lc, and judge whether the quantity of the smallest intermediate result of lcandls} is 1, if it is 1, Then use the intermediate result with the smallest l_c and l_s as the target intermediate result, or若所述l_c和l_s最小的中间结果的数量不为1，则判断所述l_c和l_s最小的中间结果中是否存在与历史目的压力感知结果的压力类型词相同的中间结果，若存在，则从所述l_c和l_s最小的中间结果中与历史目的压力感知结果的压力类型词相同的中间结果中选取一个中间结果作为所述目的中间结果，否则，若不存在，则从所述l_c和l_s最小的中间结果中选取一个中间结果作为所述目的中间结果。If the quantity of the minimum intermediate result of_lc and_ls is not 1, then judge whether there is an intermediate result identical to the pressure type word of the pressure perception result of the historical purpose in the intermediate result of the minimum_lc and_ls , if Exist, then select an intermediate result from the intermediate results that are the same as the pressure type word of the historical purpose pressure perception result from the intermediate results with the smallest l_c and l_s as the intermediate result of the purpose, otherwise, if not exist, then select an intermediate result from An intermediate result among the intermediate results with the smallest l_c and l_s is selected as the target intermediate result.4.根据权利要求1所述的疏导青少年心理压力的虚拟聊天方法，其特征在于，在所述从所述中间结果中选出目标中间结果之后，从所述目标中间结果中提取目的压力感知结果之前，还包括：4. The virtual chat method for dredging teenagers' psychological pressure according to claim 1, characterized in that, after the target intermediate result is selected from the intermediate result, the target pressure perception result is extracted from the target intermediate result Previously, also included:判断所述目标中间结果的Stress的值与历史目的压力感知结果的Stress的值是否为有压力所对应的值，若所述历史目的压力感知结果的Stress的值为有压力所对应的值，所述目标中间结果的Stress的值为无压力所对应的值，则判断所述历史目的压力感知结果的Category的值与所述目标中间结果的Category的值以及所述历史目的压力感知结果的Subcategory的值与所述目标中间结果的Subcategory的值是否相等，若相等，则将所述目标中间结果的Stress的值更新为所述历史目的压力感知结果的Stress的值，或者Judging whether the value of Stress of the target intermediate result and the value of Stress of the historical purpose pressure perception result are values corresponding to pressure, if the value of Stress of the historical purpose pressure perception result is a value corresponding to pressure, then The Stress value of the target intermediate result is the value corresponding to no stress, then judge the value of the Category of the historical target stress perception result and the value of the Category of the target intermediate result and the Subcategory of the historical target stress perception result Whether the value is equal to the value of the Subcategory of the target intermediate result, if they are equal, the value of Stress of the target intermediate result is updated to the value of Stress of the historical target pressure perception result, or若所述历史目的压力感知结果的Stress的值与所述目标中间结果的Stress的值为有压力所对应的值，则判断所述目标中间结果的Category的值与Subcategory的值是否为空值，若为空值，则将所述目标中间结果的Category的值更新为所述历史目的压力感知结果的Category的值，将所述目标中间结果的Subcategory的值更新为所述历史目的压力感知结果的Subcategory的值，否则，若所述目标中间结果的Category的值不为空值，Subcategory的值为空值，则判断所述历史目的压力感知结果的Category的值与所述目标中间结果的Category的值是否相等，若相等，则将所述目标中间结果的Subcategory的值更新为所述历史目的压力感知结果的Subcategory的值。If the value of Stress of the historical target pressure perception result and the value of Stress of the target intermediate result are values corresponding to pressure, it is judged whether the value of Category and Subcategory of the target intermediate result are null values, If it is a null value, update the Category value of the target intermediate result to the Category value of the historical target pressure perception result, and update the Subcategory value of the target intermediate result to the value of the historical target pressure perception result The value of Subcategory, otherwise, if the value of Category of the target intermediate result is not a null value, the value of Subcategory is a null value, then judge the value of the Category of the pressure perception result of the historical purpose and the Category of the target intermediate result Whether the values are equal, and if they are equal, the value of the Subcategory of the target intermediate result is updated to the value of the Subcategory of the historical target pressure perception result.5.根据权利要求1所述的疏导青少年心理压力的虚拟聊天方法，其特征在于，所述根据所述目的压力感知结果得到目标应答语句，包括：5. the virtual chat method of dredging young people's psychological pressure according to claim 1, is characterized in that, described according to described purpose pressure perception result, obtains target response sentence, comprises:确定出所述第一会话语句的类型，其中，所述类型包括疑问句和非疑问句；Determine the type of the first conversational sentence, wherein the type includes interrogative sentences and non-interrogative sentences;判断所述第一会话语句的类型是否为非疑问句以及目的压力感知结果的Stress的值是否为有压力所对应的值，若所述第一会话语句的类型为非疑问句，目的压力感知结果的Stress的值为有压力所对应的值，或者若所述第一会话语句的类型为疑问句，目的压力感知结果的Stress的值为无压力所对应的值，则查找预先建立的第一知识库，得到所述目的压力感知结果的Category的值和Subcategory的值所对应的应答语句，并将查找得到的应答语句作为所述目标应答语句，其中，所述第一知识库中保存有多条由Category的值、Subcategory的值和应答语句组成的记录，或者Determine whether the type of the first conversational sentence is a non-interrogative sentence and whether the value of Stress of the target pressure perception result is a value corresponding to pressure, if the type of the first conversational sentence is a non-interrogative sentence, the Stress of the purpose pressure perception result The value is the value corresponding to stress, or if the type of the first conversational sentence is an interrogative sentence, the value of Stress of the target pressure perception result is the value corresponding to no pressure, then search the pre-established first knowledge base to obtain The response statement corresponding to the value of Category and the value of Subcategory of the target pressure perception result, and the response statement obtained by searching is used as the target response statement, wherein the first knowledge base stores a plurality of value, the value of Subcategory and the record consisting of the answer statement, or若所述第一会话语句的类型为非疑问句，目的压力感知结果的Stress的值为无压力所对应的值，则从第二知识库中选取所述目标应答语句，其中，所述第二知识库中保存有多条应答语句，或者If the type of the first conversational sentence is a non-interrogative sentence, and the value of Stress of the target pressure perception result is a value corresponding to no pressure, then the target answer sentence is selected from the second knowledge base, wherein the second knowledge There are multiple answer statements stored in the library, or若所述第一会话语句的类型为疑问句，目的压力感知结果的Stress的值为有压力所对应的值，则从第三知识库中获取所述目标应答语句，其中，所述第三知识库中包括多个问题语句以及相应的答案。If the type of the first conversational sentence is an interrogative sentence, and the value of Stress of the target pressure perception result is a value corresponding to stress, then the target answer sentence is obtained from the third knowledge base, wherein the third knowledge base Include multiple question statements and corresponding answers.6.根据权利要求5所述的疏导青少年心理压力的虚拟聊天方法，其特征在于，所述从第三知识库中获取所述目标应答语句，包括：6. The virtual chat method for dredging young people's psychological pressure according to claim 5, wherein said obtaining said target answer sentence from the third knowledge base includes:从第三知识库中获取与所述第一会话语句最相关的问题语句，其中，所述第三知识库中还包括多个问题语句相应的答案是否被采纳的信息和相应的答案被赞同的数量信息；Obtain the question statement most relevant to the first conversational statement from the third knowledge base, wherein the third knowledge base also includes information on whether the corresponding answers of the plurality of question statements are adopted and whether the corresponding answers are approved quantity information;计算所述第一会话语句与最相关的问题语句的匹配度，判断所述匹配度是否大于预设的数值，若大于所述数值，则判断所述最相关的问题语句的答案是否被采纳，若被采纳，则从被采纳的答案中选取赞同数量最多的被采纳的答案作为所述目标应答语句，或者，若不大于所述数值或者不被采纳，则从所述第一知识库保存的鼓励话语中随机选取一个鼓励话语作为所述目标应答语句。Calculating the matching degree of the first conversational sentence and the most relevant question sentence, judging whether the matching degree is greater than a preset value, if greater than the numerical value, then judging whether the answer of the most relevant question sentence is adopted, If adopted, select the adopted answer with the largest number of approvals from the adopted answers as the target answer sentence, or, if it is not greater than the numerical value or not adopted, select the answer from the first knowledge base An encouraging utterance is randomly selected from the encouraging utterances as the target response sentence.7.根据权利要求6所述的疏导青少年心理压力的虚拟聊天方法，其特征在于，所述计算所述第一会话语句与最相关的问题语句的匹配度，包括：7. the virtual chat method of dredging young people's psychological pressure according to claim 6, is characterized in that, described calculation described first conversation sentence and the matching degree of most relevant question sentence, comprise:对所述最相关的问题语句Q_b进行中文分词，并建立第二依存关系树；Carry out Chinese word segmentation to described most relevant problem statement Q_b , and set up the second dependency tree;确定出所述第一依存关系树的有效节点对u＝(u₁,u₂)和第二依存关系树的有效节点对b＝(b₁,b₂)，其中，u＝(u₁,u₂)，u₁为所述第一依存关系树的根节点，u₂为与u₁具有直接依存关系的节点，b₁为所述第二依存关系树的根节点，b₂为与b₁具有直接依存关系的节点；Determine the valid node pair u=(u₁ ,u₂ ) of the first dependency tree and the valid node pair b=(b₁ ,b₂ ) of the second dependency tree, where u=(u₁ , u₂ ), u₁ is the root node of the first dependency tree, u₂ is a node with direct dependency with u₁ , b₁ is the root node of the second dependency tree, b₂ is the node with b₁ Nodes with direct dependencies;计算所述第一会话语句Q_u与最相关的问题语句Q_b的匹配度，Calculate the degree of matching between the first conversational statement Q_u and the most relevant question statement Q_b ,计算公式为MatchDegree(Q_u,Q_b)＝αSIM(E(Q_u),E(Q_b))+(1-α)SIM(E^*(Q_u),E^*(Q_b))，其中，MatchDegree(Q_u,Q_b)为所述第一会话语句Q_u与最相关的问题语句Q_b的匹配度，α(α∈(0,1))为常数，E(Q_u)为所述第一依存关系树的有效节点对的集合，E(Q_b)为所述第二依存关系树的有效节点对的集合，E^*(Q_u)为将所述第一依存关系树的各个节点用其义原替代后得到的依存关系树的有效节点对的集合，E^*(Q_b)为将所述第二依存关系树的各个节点用其义原替代后得到的依存关系树的有效节点对的集合，$SIM\left(E\right(Q_u),E(Q_b\left)\right)\frac{{\mathrm{\&Sigma;}}_{u\&Element;E\left(Q_u\right)}{\mathrm{\&Sigma;}}_{b\&Element;E\left(Q_b\right)}sim(u,b)}{Max\left\{\right|E\left(Q_u\right)|,|E\left(Q_b\right)\left|\right\}},$$SIM\left(E^{*}\right(Q_u),E^{*}(Q_b\left)\right)\frac{{\mathrm{\&Sigma;}}_{u^{*}\&Element;E^{*}\left(Q_u\right)}{\mathrm{\&Sigma;}}_{b^{*}\&Element;E^{*}\left(Q_b\right)}sim(u^{*},b^{*})}{Max\left\{\right|E^{*}\left(Q_u\right)|,|E^{*}\left(Q_b\right)\left|\right\}},$u^*为E^*(Q_u)的元素，b^*为E^*(Q_b)的元素，$sim(u,b)=sim\left(\right(u_1,u_2),(b_1,b_2\left)\right)=\left\{\begin{array}{cc}1& if(u_1==b_1,u_2==b_2)\\ 0.5& if(u_1==b_1,u_2\&NotEqual;b_2)\\ 0& otherwise\end{array}\right\}.$The calculation formula is MatchDegree(Q_u ,Q_b )=αSIM(E(Q_u ),E(Q_b ))+(1-α)SIM(E^* (Q_u ),E^* (Q_b )), where , MatchDegree(Q_u ,Q_b ) is the matching degree between the first conversational statement Q_u and the most relevant question statement Q_b , α(α∈(0,1)) is a constant, E(Q_u ) is all The set of valid node pairs of the first dependency tree, E(Q_b ) is the set of valid node pairs of the second dependency tree, E^* (Q_u ) is the set of each of the first dependency tree A collection of effective node pairs of the dependency tree obtained after the node is replaced with its sememe, E^* (Q_b ) is the effective node pair of the dependency tree obtained after each node of the second dependency tree is replaced with its sememe collection of node pairs,$SIm\left(\mathrm{E.}\right(Q_u),\mathrm{E.}(Q_b\left)\right)\frac{{\mathrm{\&Sigma;}}_{u\&Element;\mathrm{E.}\left(Q_u\right)}{\mathrm{\&Sigma;}}_{b\&Element;\mathrm{E.}\left(Q_b\right)}\mathrm{the\; s}im(u,b)}{max\left\{\right|\mathrm{E.}\left(Q_u\right)|,|\mathrm{E.}\left(Q_b\right)\left|\right\}},$$SIm\left({\mathrm{E.}}^{*}\right(Q_u),{\mathrm{E.}}^{*}(Q_b\left)\right)\frac{{\mathrm{\&Sigma;}}_{u^{*}\&Element;{\mathrm{E.}}^{*}\left(Q_u\right)}{\mathrm{\&Sigma;}}_{b^{*}\&Element;{\mathrm{E.}}^{*}\left(Q_b\right)}\mathrm{the\; s}im(u^{*},b^{*})}{max\left\{\right|{\mathrm{E.}}^{*}\left(Q_u\right)|,|{\mathrm{E.}}^{*}\left(Q_b\right)\left|\right\}},$ u^* is an element of E^* (Q_u ), b^* is an element of E^* (Q_b ),$\mathrm{the\; s}im(u,b)=\mathrm{the\; s}im\left(\right(u_1,u_2),(b_1,b_2\left)\right)=\left\{\begin{array}{cc}1& if(u_1==b_1,u_2==b_2)\\ 0.5& if(u_1==b_1,u_2\&NotEqual;b_2)\\ 0& otherwi\mathrm{the\; s}e\end{array}\right\}.$8.根据权利要求1所述的疏导青少年心理压力的虚拟聊天方法，其特征在于，在所述获取用户的第一会话语句之前，还包括：8. The virtual chat method for dredging young people's psychological pressure according to claim 1, characterized in that, before the acquisition of the user's first conversational sentence, it also includes:接收用户的直接登录指令，并将第一应答语句反馈给用户，或者Receive the user's direct login instruction and feed back the first response statement to the user, or接收用户的登录指令，并在用户授权连接微博时，获取用户的原创微博，根据所述原创微博判断用户是否有压力情绪，若有，则将第二应答语句反馈给用户，否则，则将第三应答语句反馈给用户。Receive the user's login instruction, and when the user authorizes to connect to the microblog, obtain the user's original microblog, judge whether the user is stressed according to the original microblog, and if so, feed back the second answer statement to the user, otherwise, then Feedback the third answer sentence to the user.9.根据权利要求8所述的疏导青少年心理压力的虚拟聊天方法，其特征在于，在所述获取用户的第一会话语句之前，还包括：9. The virtual chat method for dredging young people's psychological pressure according to claim 8, characterized in that, before the acquisition of the user's first conversational sentence, it also includes:在将所述第一应答语句、第二应答语句或者第三应答语句反馈给用户之后，获取用户的第二会话语句，对所述第二会话语句进行压力情绪和压力类型检测；After feeding back the first answer sentence, the second answer sentence or the third answer sentence to the user, obtain the user's second conversation sentence, and perform stress emotion and stress type detection on the second conversation sentence;若所述第二会话语句中检测出压力情绪，未检测出压力类型，则对所述原创微博进行压力检测，并基于所述压力检测结果将第四应答语句反馈给用户，或者If the stress emotion is detected in the second conversation sentence, but the pressure type is not detected, the stress detection is performed on the original microblog, and the fourth response sentence is fed back to the user based on the stress detection result, or若所述第二会话语句中检测出压力情绪和压力类型或者未检测出压力情绪，则执行步骤S1。If the stress emotion and stress type are detected in the second conversation sentence or the stress emotion is not detected, step S1 is executed.10.一种疏导青少年心理压力的虚拟聊天装置，其特征在于，包括：10. A virtual chat device for relieving psychological pressure of teenagers, characterized in that it comprises:建立单元，用于获取用户的第一会话语句，对所述第一会话语句进行中文分词，并建立第一依存关系树，其中，所述第一依存关系树的根节点是所述第一会话语句的中心词，各个子节点为所述进行中文分词得到的词，每一条边表示该边相应的子节点的依存关系；The establishment unit is used to obtain the user's first conversation sentence, perform Chinese word segmentation on the first conversation sentence, and establish a first dependency tree, wherein the root node of the first dependency tree is the first conversation The central word of the sentence, each child node is the word obtained by performing Chinese word segmentation, and each edge represents the dependency relationship of the corresponding child nodes of the edge;计算单元，通过预先建立的压力词库对所述第一依存关系树的各个节点进行标注，并计算标注后的压力类型节点与情绪词节点之间的路径长度以及压力子类型节点与情绪词节点之间的路径长度得到中间结果，其中，所述压力词库包括否定词以及其依存的正面情绪词、压力情绪词、压力类型词、压力类型词下的压力子类型词和属于压力子类型词所对应的压力子类型范畴的描述词，所述中间结果的格式为(Stress，Category：l_c，Subcategory：l_s)，Stress表示是否有压力，Category表示压力类型词，l_c表示压力类型词与情绪词节点之间的路径长度，Subcategory表示压力子类型词，l_s表示压力子类型词与情绪词节点之间的路径长度；The calculation unit is used to mark each node of the first dependency tree through the pre-established pressure lexicon, and calculate the path length between the marked pressure type node and the emotional word node, as well as the pressure subtype node and the emotional word node The path length between obtains the intermediate result, wherein, described stress lexicon comprises negative word and its dependent positive emotion word, stress emotion word, pressure type word, pressure subtype word under pressure type word and belongs to pressure subtype word The descriptor of the corresponding stress subtype category, the format of the intermediate result is (Stress, Category: l_c , Subcategory: l_s ), Stress represents whether there is stress, Category represents the stress type word, and l_c represents the stress type word and the path length between the emotion word node, Subcategory represents the pressure subtype word, and l_s represents the path length between the pressure subtype word and the emotion word node;提取单元，用于从所述中间结果中选出目标中间结果，从所述目标中间结果中提取目的压力感知结果，其中，所述目的压力感知结果的格式为(Stress，Category，Subcategory)；An extraction unit, configured to select a target intermediate result from the intermediate results, and extract a target stress perception result from the target intermediate result, wherein the format of the target stress perception result is (Stress, Category, Subcategory);获取单元，用于根据所述目的压力感知结果得到目标应答语句，并反馈给用户。An acquisition unit, configured to obtain a target response sentence according to the target pressure perception result, and give feedback to the user.