CN101436206B - Tourism request-answer system answer abstracting method based on ontology reasoning - Google Patents

Tourism request-answer system answer abstracting method based on ontology reasoning Download PDF

Info

Publication number
CN101436206B
CN101436206B CN 200810233734 CN200810233734A CN101436206B CN 101436206 B CN101436206 B CN 101436206B CN 200810233734 CN200810233734 CN 200810233734 CN 200810233734 A CN200810233734 A CN 200810233734A CN 101436206 B CN101436206 B CN 101436206B
Authority
CN
China
Prior art keywords
concept
answer
ontology
tourism
question
Prior art date
Application number
CN 200810233734
Other languages
Chinese (zh)
Other versions
CN101436206A (en
Inventor
余正涛
张宜浩
张志坤
毛存礼
郭剑毅
龚华明
Original Assignee
昆明理工大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 昆明理工大学 filed Critical 昆明理工大学
Priority to CN 200810233734 priority Critical patent/CN101436206B/en
Publication of CN101436206A publication Critical patent/CN101436206A/en
Application granted granted Critical
Publication of CN101436206B publication Critical patent/CN101436206B/en

Links

Abstract

The invention relates to an answer extraction method for a tourism question and answer system based on ontology reasoning, which belongs to the field of artificial intelligence. The answer extraction method for the tourism question and answer system based on ontology reasoning is characterized in that: firstly, semantic rules in the field are defined, an artificial ontology knowledge base is constructed, and question sentences of users are analyzed; secondly, answer extraction is performed by combing semantic rule-based reasoning and information retrieval, instead of simple matching; and thirdly, corresponding answer extraction algorithms are designed according to different question sentence types. The invention provides the answer extraction method for the tourism question and answer system based on ontology reasoning, which introduces the ontology concept into construction of the knowledge base of the question and answer system, uses an OWL (Ontology Web Language) ontology descriptive language to clearly and definitely represent concepts, attributes and relations in the field of tourism, and more effectively organizes knowledge. In an open test, the precision of answers of the question and answer system based on ontology reasoning to 1346 natural language questions of the users reaches 81.35 percent, and the recall rate reaches 90.49 percent.

Description

基于本体推理的旅游问答系统答案抽取方法技术领域 Ontology-based reasoning system of Forum answer technical field extraction method

[0001] 本发明涉及一种基于本体推理的旅游问答系统答案抽取方法。 [0001] The present invention relates to an ontology reasoning system answers the forum based extraction method. 属人工智能领域。 It belongs to the field of artificial intelligence. 背景技术 Background technique

[0002] 自动问答系统,又称QA(Queslion Answering)系统,是一种智能新技术,它采用自然语言处理技术,一方面完成对用户疑问的分析处理:另一方面完成正确答案的生成,让人们在杂乱无章的网络世界中快速、准确地获得自己想要的信息。 [0002] QA system, also known as QA (Queslion Answering) system, an intelligent new technology, which uses natural language processing technology, on the one hand to complete the analysis and processing of the user in question: on the other hand is completed to generate the correct answer, so quickly and accurately obtain the information they want in the online world in chaotic. 在现阶段,要让计算机完全理解人类语言还非常困难,但是对于特定的领域,采用针对性的方法,已经开发出许多成功的应用案例。 At this stage, let the computer completely understand human language is still very difficult, but for specific areas, using targeted methods have been developed many successful application cases.

[0003] 在受限领域自动问答系统中,答案抽取部分是一个难点,关系到整个问答系统的最终效果。 [0003] In the field of automatic question answering system is limited, answer extraction is a difficult part, related to the final effect of the whole question answering system. 目前的问答系统答案抽取主要分为聊天机器人问答系统答案抽取、基于Web的开放式问答系统答案抽取、基于知识的问答系统答案抽取。 The current quiz answer extraction system is divided into bot quiz answer extraction system, a Web-based open question answering system answer extraction, extraction-based question answering system answers knowledge. [0004] 聚天机器人问答系统在答案抽取的时候采用模式匹配的方法,来寻找问题最合适的答案。 [0004] Poly days Q robot system uses a pattern matching extraction when the answer to the problem of finding the most appropriate answer. 其特点是在与用户的交谈过程中,基于谈话技巧和程序技巧,而不是根据常识。 Its characteristics are in the process of talking with the user, based on the interview techniques and skills programs, rather than common sense. 在它们的对话库中,可以存放多个句型、模板,但几乎没有常识库。 In their conversations library, you can store multiple sentences, templates, but almost no knowledge base. 这种答案抽取方法由于缺乏知识,所以其实际用途不大。 This answer extraction method due to lack of knowledge, so little actual use.

[0005] 而基于Web的开放式问答系统先从Web上检索一些相关文档,对相关文档采取答案抽取技术抽取答案。 [0005] and start the Web to retrieve some documents related to Web-based system of open questions and answers, to take the relevant documentation answer extraction technology to extract the answer. 但是,目前的基于Web问答系统大多局限在某个特定领域或者特定范围之内,能够回答的问题类型也比较简单,真正的面向Web开放域的问答系统的正确率和精确性都不高,还不能提供良好的商业服务。 However, the current Web-based Q & A system mostly confined to a particular area or a particular range, able to answer the types of questions are relatively simple, the true accuracy and precision of question answering system for the open Web domain is not high, but also We can not provide good business service.

[0006] 基于知识的问答系统一般是受限领域问答系统,它包含自然语言界面的专家系统、基于受限语言的数据库查询系统、基于FAQ的问答系统、基于本体的问答系统。 [0006] Q & A knowledge-based systems are generally limited field of QA system, which includes natural language expert system interface, database query language-based systems are limited, based on the FAQ's answering system, ontology-based question and answer system. 自然语言界面的专家系统一般采用各种专家系统语言=PROLOG语言、ALLTALK语言、LISP语言等来分析回答用户的疑问,给出回答,现有的专家系统一般知识库和推理、回答机制不分离,它们按知识在专家系统语言基础上开发程序,使用范围小,可移植性不高。 Expert systems generally use natural language interface of various expert system language = PROLOG language, ALLTALK language, LISP language to analyze the answer user questions, give answers, existing expert system knowledge base and general reasoning, the answer is not separation mechanism, them in the development of knowledge in expert system on the basis of language program, use a small range, portability is not high. 基于受限语言的数据库查询系统将问句转换为数据库的SQL语句,通过SQL语句在系统数据库中查询答案,这需要一个大数据库的支持,数据库的构建标准很难确定,而且用数据库方式不太适合组织领域知识库。 Database query language-based systems are limited to converting SQL statement question database through SQL statements answer queries in the system database, which needs the support of a large database, construct a standard database is difficult to determine, but the database is not the way domain knowledge base for your organization. 基于FAQ的问答系统先计算用户问句和FAQ知识库中问题的相似度,从而找到FAQ知识库中与用户查询最为相似的问题,然后把此问题对应的相关答案直接提交给用户,基于FAQ的问答系统回答范围有限,它能回答的内容基本上是问答对所包含的内容,很难用问答对来组织领域内所有的知识。 Submitted directly based FAQ system to calculate the similarity of questions and answers user questions and problems FAQ knowledge base, FAQ knowledge base in order to find the user's query is most similar problems, then the answer to this question related to the corresponding user, based on the FAQ Q & a system limited the scope of the answer, it is basically a question and answer content to answer all of the knowledge in the field of organizing content contained difficult question and answer pairs. 基于本体的问答系统现在正处于研究阶段,怎样更好的利用本体来进行答案抽取是一个热门话题。 Ontology-based Q & A system is now in the research phase, how to better use the body to answer extraction is a hot topic.

[0007] 同时,目前的问答系统缺乏推理能力,推理系统缺乏自然语言理解能力。 [0007] Meanwhile, the current system of questions and answers lack of reasoning ability, reasoning systems lack the ability to understand natural language. 正是这个问题困扰着大型知识库系统的建设,也使花费巨大的人力物力建立起来的知识库系统难以面向公众开展达到一定质量的知识服务。 It is this problem plaguing the construction of large knowledge base, but also to spend a huge human and material resources built up a knowledge base system is difficult for the public to carry out a certain quality of service knowledge.

[0008] 本体(ontology)原是哲学研究中发展出来的一个概念,研究客观事物存在的本质和组成。 Was originally a concept study philosophy [0008] ontology (ontology) are developed, the nature of things, the existence of objective research and composition. 本体在哲学定义上的主要特点在于本体是关于世界某个方面的一个特定的分类体系,这个体系不依赖任何特定的语言。 The main features of the body on the philosophy that the body is defined by a specific classification system on some aspect of the world, this system does not depend on any particular language. 近年来,随着信息科学的飞速发展,本体逐渐用于识识工程和信息科学等领域之中。 In recent years, with the rapid development of information science, the body gradually into the field of knowledge for knowledge engineering and information science.

[0009] 本体在国外已经成为研究热点,在多个领域出现了具体应用,其研究集中在知识工程,本体工程、信息组织与检索和语义Web等方面。 [0009] body in foreign countries has become a hot topic, there have been specific applications in many fields, the research focused on knowledge engineering, ontology engineering, information organization and retrieval and semantic Web and so on. 比较著名的通用本体研究包括CYC项目和Chan2drsekaran等的关于任务的问题求解方法本体的研究。 More well-known general body of research, including CYC project and Chan2drsekaran and other questions about the task of solving the bulk of research methods. 前者是美国的微电子与计算机技术有限公司的研究顶目,目标是开发本体,进行常识推理,目前已经发展成为一个庞大的常识系统;后者则是研究可共享问题的求解方法,与领域无关的推理方法。 The former is the study of the top head of the US Microelectronics and Computer Technology Co., Ltd. goal is to develop a body, a commonsense reasoning, it has now developed into a vast knowledge of the system; the latter is the study of methods to solve shared problems, regardless of the field method of reasoning. 比较著名的领域本体研究包括爱丁堡大学的企业项目和多伦多大学的虚拟企业项目。 More well-known body of research areas, including enterprise project at the University of Edinburgh and the University of Toronto Virtual Enterprise project.

[0010] 国内对于本体的研究已有很多年的时间了,比较有影响的有中科院数学所陆汝钤研究员领导的常识知识的实用性研究,中科院计算技术研究所曹存根研究员主持的大规模知识系统的研究。 [0010] domestic research for many years a time for the body, there are more influential studies of common sense practical knowledge of researchers led by Lu Ruqian Academy of Mathematics, Institute of Computing Technology Chinese Academy of Sciences researcher Cao stub hosted large-scale knowledge systems the study. 以及中科院数学研究所金芝研究员研究的基于本体的软件需求获取方法、等。 And access to software requirements based on ontology approach Kim Ji-researcher of Chinese Academy of Sciences Research Institute of Mathematics, and so on. 比较有名的通用本体构建研究包括中科院计算技术研究所的大规模知识系统研究和中科院数学研究所的常识知识库研究。 Construction of the more famous general body of knowledge, including knowledge of large-scale systems technology research institute and the Institute of Mathematics of the Chinese Academy of Sciences Institute of Computing knowledge base.

发明内容 SUMMARY

[0011] 本发明实现了一种基于本体推理的问答系统答案抽取方法,该方法将答案抽取过程分为三步,第一步定义领域中的语义规则,然后将基于语义规则推理与信息检索相结合进行答案抽取,最后再根据不同的问句类型设计相应的答案抽取算法,提高答案抽取的准确率和召回率。 [0011] The present invention achieves a quiz answer extraction method based on the system ontology reasoning, the answer extraction method three-step process, the first step in semantic rules defined in the art, and the inference rules and semantic information retrieval phase carried out in conjunction with answer extraction, and finally design the appropriate answer extraction algorithm to improve the answer extraction of precision and recall rates depending on the type of questions.

[0012] 本发明目的在于提出利用本体建立知识库,再定义领域的语义规则,并对用户问句意图进行分类分析,最后利用基于语义规则的推理和信息检索相结合的答案抽取方法从本体知识中进行答案的抽取,在开放测试中,基于本体推理的问答系统对于用户的1346条自然语言提问的回答,准确率达到了81. 35%,召回率达到了90. 49%,取得了良好的效果。 [0012] The object of the present invention is to provide the use of ontology knowledge, and then define the field of semantic rules, and the user intent question classification analysis, using the final answer extraction reasoning based information retrieval and semantic rules from combining ontology the decimation of the answer, in the open test, the question answering system based on ontology reasoning for the 1346 user's natural language question answering, accurate rate reached 81.35 percent, the recall rate reached 90.49 percent, and achieved good effect.

[0013] 一种基于本体推理的旅游问答系统答案抽取方法,其特征在于:第一步以旅游概念作为该旅游本体的顶层,以领域概念、领域属性和领域关系作为本体知识库的构建资源,构成旅游本体的中层,针对每一要素再继续划分,产生底层:然后对知识库进行实例的扩充,构建旅游本体知识库,并对用户问句进行分析;第二步将基于语义规则推理与信息检索相结合进行答案抽取;最后再根据不同的问句类型设计相应的答案抽取算法; [0013] A method for answer extraction system Forum based ontology reasoning, wherein: the first step to the concept of travel as the top of the travel body, the relationship to the field of the art concepts, properties and resources the art as building ontology knowledge base, middle constitute tourism body, for each element and then continue to divide, produce the bottom: and then expand the knowledge base instance, to build tourism ontologies, and analyzes user questions; second step of semantic reasoning and rule-based information retrieving answer extraction performed in combination; and finally extraction algorithm design corresponding answer questions based on different types;

[0014] 该方法具有包括有: [0014] The method comprises:

[0015] (I)人工定义旅游领域中的概念、属性和关系,并构建旅游领域本体知识库,最后再对本体的一致性进行检验; [0015] (I) concepts, attributes and relationships defined in the artificial field of tourism, and build tourism domain ontology knowledge base, and finally to the consistency of the inspection body;

[0016] (2)利用步骤(I)的本体知识库中的语义信息对用户问句进行语义消歧; Ontologies semantic information [0016] (2) using the step (I) to the user in question semantic disambiguation;

[0017] (3)人工自定义旅游领域中的语义规则;自定义语义规则如下: [0017] (3) custom artificial semantic rule in the field of tourism; custom semantic rule as follows:

[0018] 【Rule I :K(x, y), A(x, z)_>A(y, z)】 [0018] [Rule I: K (x, y), A (x, z) _> A (y, z)]

[0019] 代表上位概念有的属性,下位概念也有该属性; [0019] Representative properties of some superordinate concept, a subordinate concept also has the property;

[0020] 【Rule 2 :S(x, y), A(x, z)_>A(y, z)】 [0020] [Rule 2: S (x, y), A (x, z) _> A (y, z)]

[0021] 代表某个概念有的属性,其相似概念也有该属性; [0021] represents a concept and some properties, which also have the property similar concepts;

[0022] [Rule 3 :K(x, y), x(R) = z->y (R) = z][0023] 代表上位概念X和概念z有角色关系R,则其下位概念y和概念z也有角色关系R : [0022] [Rule 3: K (x, y), x (R) = z-> y (R) = z] [0023] Representative generic term X and concepts z has a role relationships R, it subordinate concept y and The concept also has a role relationship z R:

[0024] 【Rule 4 :S(x, y), x(R) = z — y (R) = z】 [0024] [Rule 4: S (x, y), x (R) = z - y (R) = z}

[0025] 代表某个概念X和概念z有角色关系R,则其相似概念y和概念z也有角色关系R : [0025] represents a concept X and z have the concept of character relation R, the concept of which is similar to the concept of y and z have the role relationship R:

[0026] 其中上述的K(x,y)表示X是y的上位概念,S(x,y)表示x和y是相似概念,A(x,z)表示z是X的属性,X (R) =z表示概念X和概念z有角色关系R : [0026] wherein the above-described K (x, y) represents the X is a superordinate concept of y, S (x, y) represents the x and y are similar concepts, A (x, z) represents a z is the property of X, X (R ) = z z concepts and conceptual X have role relation R:

[0027] (4)基于步骤(2)的问句分析结果,采用基于步骤(3)中的语义规则的推理和信息检索相结合的方法在步骤(I)的本体知识库中抽取答案; [0027] (4) based on the step (2) analysis of the question, steps of a method based information retrieval and semantic reasoning rule (3) in combination ontologies answer extraction step (I) of;

[0028] (5)根据步骤(2)中的不同的问句类型,设计相应的答案抽取算法,在不降低答案抽取速度的基础上,提高系统的响应率和召回率。 [0028] (5) according to step a different question types (2), the corresponding answer extraction algorithm design, without reducing the speed of answer extraction, to improve system response and recall.

[0029] 步骤⑴中定义了旅游领域中的概念、属性和关系,构建的领域本体知识库(云南旅游)。 [0029] Step defines the concept, properties and relationships in the field of tourism ⑴ construct domain ontologies (in Travel).

[0030] 步骤⑵中利用本体知识库中的语义信息对用户问句进行语义消歧。 [0030] Step ⑵ user in question semantic disambiguation semantic ontology knowledge base information.

[0031] 步骤(3)中自定义的旅游领域语义规则。 [0031] Step semantic rules art custom travel (3).

[0032] 步骧⑷中的规则推理与信息检索相结合的答案抽取方法。 [0032] answer extraction method of inference rules and information retrieval Xiang ⑷ step of combining.

[0033] 步骤(5)中的根据不同问句类型设计相应的答案抽取算法。 [0033] Step design appropriate answers questions depending on the type of extraction algorithm (5).

[0034] 本发明将本体的思想引入问答系统知识库的构建,把旅游领域中的概念、属性和关系用OWL (Ontology Web Language)本体描述语言清晰明确地表示出来,更加有效地组织知识。 [0034] The invention will be introduced into the body of thought answering system to build the knowledge base, the concept of tourism fields, attributes, and relationships with OWL (Ontology Web Language) ontology language clear that out, more effective organization of knowledge.

[0035] 本发明本体描述某个领域或更广范围内的概念以及概念之间的关系,而这些概念和关系在共享的范围内具有大家认可的、明确的、唯一的定义。 Body [0035] The present invention describes the relationship between the concept and the concept of a field or a wider range, and these concepts and relationships have all recognized, clearly, only defined within a range shared. 在受限领域问答系统中采用本体知识库,可以更好的表示知识之间的内在关系,知识的组织更加合理,减少冗余存储,提高答案抽取的准确率和召回率。 The use of ontologies in a restricted area of ​​question answering system, you can better represent the intrinsic relationship between knowledge, organizational knowledge is more reasonable, reduce redundant storage, improve the accuracy and recall the answer extraction. 在开放测试中,基于本体推理的问答系统对于用户的1346条自然语言提问的回答,准确率达到了81. 35%,召回率达到了90. 49%。 In the open test, question answering system based on ontology reasoning for the 1346 user's natural language question answering, accurate rate reached 81.35 percent, the recall rate reached 90.49 percent.

附图说明 BRIEF DESCRIPTION

[0036] 图I是本发明中所定义的旅游本体类结构图。 [0036] Figure I is a configuration diagram of travel ontology class defined in the present invention.

[0037] 图2是本发明提出的基于本体推理的问答系统答案抽取方法的流程图。 [0037] FIG 2 is a flowchart of a method based on extraction ontology reasoning system quiz answer proposed by the present invention.

具体实施方式 Detailed ways

[0038] 我们构建的本体知识库中,共收集了2380条云南旅游本体实例,过程为: [0038] we constructed ontology knowledge base, collected a total of 2380 Yunnan tourism ontology instances, the process is:

[0039] 一、定义本体类结构 [0039] First, the structure ontology class definitions

[0040] 本体知识库是问答系统的大脑,其优劣直接关系到后续的问句分析以及答案抽取的效率以及整个系统的性能。 [0040] Knowledge of the body is the brain of the system Q, which is directly related to the merits of the subsequent analysis and question and answer extraction efficiency of the overall system performance. 因此,在建立领域本体知识库的过程中,我们从领域本体所涉及的范围,应用目的等方面来考虑。 Therefore, in the process of establishing domain ontology knowledge base, we have to consider in terms of the range of areas of the body involved, applications and other purposes. 设计一个本体的过程一般包括:确定应用范围,确定本体中的概念,属性,确定本体中概念与概念之间以及属性与属性之间的关系,对本体进行编码,对本体的能力进行评估。 A body design process generally comprises: determining the scope of application, the concept of determining the body, attribute, determine the relationship between the concept and the concept between the body and the properties and attributes of the body is encoded, the ability of the body to be evaluated. 这样就可以生成一个较完整的知识库。 This will generate a more complete knowledge base. 实验采用自顶向下的方法,从领域中概括出主要概念,并逐步细化,建立子类。 Experiments using top-down approach, summarized the main concepts from the field, and gradually refined subclassing. 分析旅游所涉及的小吃、住宿、旅行、购物、娱乐以及风土民情等要素。 Analysis of factors snacks, accommodation, travel, shopping, entertainment and local customs and other travel involved. 共定义19个领域概念。 19 areas were defined concept. 图I所示为旅游本体类的信息和结构。 Figure I is a body of information and tourist class structure shown in FIG. 首先,以旅游概念作为该旅游本体的Top-level,再粗粒度的将其分为特色食物、住宿、交通方式、风景名胜、地理位置、特产、娱乐活动、少数民族、民族风情等19类,这些构成了旅游本体的Middle-level。 First, the concept of tourism as a Top-level tour of the body, and then will be divided into coarse-grained features food, accommodation, transportation, attractions, location, specialty, recreational activities, minority, ethnic customs and other 19 categories, these constitute the Middle-level tourism body. 针对每一要素再继续划分,就产生了bottom-level。 For each element and then continue to divide, they produce a bottom-level.

[0041] 二、定义本体中的属性 [0041] Second, the definition of the properties of the body

[0042] 仅有类对很多问题都不能给出回答,因此还需要定义概念和概念间的内部联系。 [0042] just like to give a lot of problems can not be answered, it is also necessary internal relations between the concepts and definitions. 这里所指的联系可分为两种:一种是概念自身的属性,称为“内在属性”,如概念“民族服饰”的颜色这种属性,这一类属性通常连接一个概念和一个值,在OWL中,这种属性被表示为DatatypeProperty。 Information referred to herein can be divided into two types: one is a conceptual own attributes, referred to as "intrinsic property", such as the concept of "national dress" colors such properties, the properties are usually connected to a concept and a value, in OWL, this property is represented as DatatypeProperty. 内在属性具有通用性,也就是说该类对应的所有实例都具有这种属性,并且这种属性通常能向下传递,即如果各类具有一个内在属性,那么它的所有子类都继承了这种属性。 Intrinsic properties versatile, which means that all instances of the class that have corresponding properties, and such properties generally can pass downwardly, i.e., if all kinds having an inherent attribute, then all its subclasses that inherit kinds of attributes. 这样也就要求在属性建模的过程中,一个属性应该为拥有该属性的最大类所拥有。 This also requires the modeling process attributes, one attribute should have the largest class of the property owned.

[0043] 另一类属性称为“外在属性”,也有的文献直接称之为“关系”,通常用于连接概念间的实例,如概念“风景名胜”的一个外在属性“Locate”连接了概念“地理位置”,表明对于一对分别来自这两个概念的实例来说,可能会存在“Locate”这个关系。 [0043] Another attribute is called "external Properties", also direct some literature referred to as "relationship" between commonly connected Examples concepts, such as the concept of "scenic" attribute of an external "the Locate" connection the concept of "location", indicating that the pair of instances from each of these two concepts, it may exist "Locate" this relationship. “外在属性”在OWL语言中用owl :0bjectProperty定义,并可以用rdfs :domain和rdfs :range指明它的定义域和作用域。 "Extrinsic properties" in OWL language using owl: 0bjectProperty defined, and may be used rdfs: domain and rdfs: range specified domain and its scope. 还可以将一个属性定义为某个已有属性的子属性。 You can also define a property for a sub-attributes of an existing property.

[0044] 综合考虑旅游领域中概念的特性,共定义了34个内在属性,23个外在属性。 [0044] Considering the characteristics in the field of tourism concepts, it defines a total of 34 internal properties, 23 external attributes.

[0045] 三、定义属性特性和属性约束 [0045] Third, the defined attributes and attribute constraint characteristic

[0046] a、属性特性 [0046] a, characteristic properties

[0047] OffL属性拥有可传递、函数和逆关系等特性,还支持对属性取值的基数约束,从而增强了对属性的推理能力。 [0047] OffL property owned properties can be passed, function and inverse relations, but also supports cardinality constraints on property values, thereby enhancing the ability of reasoning on the property. 下面用P(x,y)表示X是P属性值为1,也可理解为X和y之间存在P关系。 Below is represented by P (x, y) X is an attribute value P 1, P is also understood that there is a relationship between X and y.

[0048] (al)传递属性(TransitiveProperty):对于任意传递属性P,如果存在P (x, y)和P(y,z),则有Ρ(χ, ζ) ο [0048] (al) transfer attribute (TransitiveProperty): For any transfer property P, if present, P (x, y) and P (y, z), there Ρ (χ, ζ) ο

[0049] (a2)对称属性(SymmetricProperty):对于任意对称属性P,如果存在P (x, y),则有P(y,X)。 [0049] (a2) the symmetry property (SymmetricProperty): For any symmetric property P, if present, P (x, y), there is P (y, X).

[0050] (a3)函数属性(FunctionalProperty):对于任意函数属性P,如果存在P (x, y)和P (X, ζ),则有y和ζ必是同一个个体或文字。 [0050] (a3) ​​function property (FunctionalProperty): For any function attributes P, if present, P (x, y) and P (X, ζ), and [zeta] y there must be the same individual or a text. 可以简单的在属性定义中用rdf :type属性声明属性具有函数特性,如: Simply using the attribute definition rdf: type attribute having a function characteristic declared attributes, such as:

[0051] <owl:ObjectProperty rdf:ID = ” IocateWhere”> [0051] <owl: ObjectProperty rdf: ID = "IocateWhere">

[0052] <rdf: type rdf: resource =,,owl: FunctionalProperty,,/> [0052] <rdf: type rdf: resource = ,, owl: FunctionalProperty ,, />

[0053] </owl:0bjectProperty> [0053] </ owl: 0bjectProperty>

[0054] 声明IocateWhere是函数属性,即任何酒店所在的地理位置都是唯一的, [0054] Statement IocateWhere is a function of property, that is, any location where the hotel is unique,

[0055] (a4)逆属性(InverseFunctional): 一个属性A被称为另一个属性B的逆,如果对任意个体X,y间有A关系,当且仅当y,X间有B关系。 [0055] (a4) an inverse property (InverseFunctional): A is a property called an inverse B of another attribute, if, for any individual in an A relationship between X y, if and only if there is the relationship between B y, X. 注:数据类型属性(DatatypeProperty)没有逆属性。 Note: The data type attributes (DatatypeProperty) no inverse property. 如下面声明hasSights是IocateWhere属性的逆: The following is a statement as hasSights inverse IocateWhere attributes:

[0056] <owl: ObjectProperty rdf: ID = ” hasSights,,> [0056] <owl: ObjectProperty rdf: ID = "hasSights ,,>

[0057] <owl: inverseOf rdf: resource =,,#locateWhere,,/>[0058] 〈/owl :0bjectProperty> [0057] <owl: inverseOf rdf: resource = ,, # locateWhere ,, /> [0058] </ owl: 0bjectProperty>

[0059] OffL DL不允许将一个数据类型属性声明为传递属性、对称属性或反函数属性。 [0059] OffL DL does not allow a data transfer is declared attribute type attribute, property or anti-symmetric function attribute. 由于抽象语法是对应OWL DL的,因此其中数据类型属性只允许声明是函数属性。 Since the corresponding abstract syntax OWL DL, and thus where only the data type attribute is declared function attribute. 根据描述逻辑理论,同时有函数性和传递性的属性会造成推理问题不可判定。 The description logics, while the properties of the transfer function and reasoning not cause determination. 因此OWL DL对属性特性的使用做出一定的限制,任意一个属性都不能同时是传递属性和(反)函数属性。 OWL DL thus make certain restrictions on the use of characteristic properties, any attributes can not be simultaneously transmitted and attributes (inverse) function attributes.

[0060] b、属性约束 [0060] b, attribute constraint

[0061] 前面的属性特性主要是对属性的全局定义域和值域的约束,但很多时候属性的值域是根据上下文变化的。 [0061] The characteristic properties of the foregoing main constraints on the global attributes of the domain and range, but often range attribute is changed depending on the context. 这些属性约束主要包括: These attributes constraints include:

[0062] (bl)owl :allValuesFrom属性约束要求对于每一个有指定属性的类实例,该属性的值必须是有owl :allValuesFrom从句指定的类的实例。 [0062] (bl) owl: allValuesFrom properties constraints require that for each instance of a class with a specified attribute value of this attribute must be a owl: allValuesFrom clause instance of the class specified.

[0063] (b2) owl :someValuesFrom 属性约束与owl :alIValuesFrom相似,它要求类实例至少有一个指定属性的值是指定的类的实例。 [0063] (b2) owl: someValuesFrom attribute constraint and owl: alIValuesFrom similar, it requires at least the value of a class instance attribute is specified instance of the specified class. 如: Such as:

[0064] 〈owl: Restriction〉 [0064] <owl: Restriction>

[0065] <owl: onProperty rdf:resoure =,,#locateWhere,,> [0065] <owl: onProperty rdf: resoure = ,, # locateWhere ,,>

[0066] <owl :hasValue rdf: resource =,,# 地理位置”〉 [0066] <owl: hasValue rdf: resource = ,, # Location ">

[0067] 〈/owl !Restriction〉 [0067] </ owl! Restriction>

[0068] 定义一个匿名类,包含所有至少有一个IocateWhere属性值为“地理位置”类的个体。 [0068] define an anonymous class, comprising at least one individual in all IocateWhere attribute value of "location" category.

[0069] 四、本体一致性的检验 [0069] Fourth, test the consistency of the body

[0070] 本体的一致性检验就是要确保本体包含的所有知识之间没有矛盾,其各组成部分构成一个协调的整体。 Consistency test [0070] body is to ensure that there is no contradiction between the body contains all the knowledge of its individual component parts of a coherent whole. 此部分工作主要从类间关系的一致性和基于公理的知识一致性两个方面着手对旅游本体的一致性进行检验。 This consistency between the main part of the work relationship and working class knowledge-based consistency axiom of two aspects of consistency tourism body is examined from.

[0071] 本体的一致性检验主要通过检验概念的可满足性来实现。 Consistency checking [0071] the body primarily through the proof of concept satisfiability. 检验一个概念的可满足性实际是看是否有解释使得这个概念成立。 A test of the concept can actually meet to see if this interpretation makes the concept of the establishment. 对一个概念C,如果存在一个解释I使得C是非空的,则称概念C是可满足的,否则是不可满足的。 C for a concept, so that if there is an explanation I C is not empty, the concept C is said to be satisfied, otherwise not satisfied. 主要通过以下五类推理来实现: Primarily through the following five types of reasoning:

[0072] (a)类(概念)——实例关系推理:给定知识库K,C是K中的一个类(概念),i是K中的一个个体,可对以下类与实例的关系进行推理:判断一个个体时候是C的一个实例:判断在K中C的所有实例:判断在K中i是那些类的实例;判断两个实例之间的关系或判断与某个实例有特定关系的实例。 [0072] (a) class (concept) - Example Inference: Given knowledge K, C K is a class of (concept), i is a K in the individual, may be made to the following classes and examples of the relationship inference: determining when an individual is one example of C: Analyzing all instances of K, C: the K i is determined in instances of those classes; determination or determining the relationship between the two instances of a specific example of the relationship between instance.

[0073] (b)类(概念)的关系推理:给定类C和D,判断它们之间的关系,主要有子类关系、成员关系以及整体与部分的关系等等。 [0073] (b) class (concept) of Inference: given class C and D, the relationship between them is determined, there are a subclass relationship, and a membership relationship with the whole portion or the like.

[0074] (c)在类的体系结构中进行推理:给定类C,返回在K中C的所有或相关的超类。 [0074] (c) In the reasoning architecture class: C given class, or to return all relevant super class C in K. 或者在K中C的所有或相关的子类。 Or all or a relevant sub-class C in K.

[0075] (d)类的满足性推理:给定一个类C,判断是否C在K中是可满足的(即一致的)。 [0075] (d) satisfy the class reasoning: Given a category C, C is determined whether (i.e. uniform) may be satisfied in K.

[0076] (e)基于属性的推理:属性与类(实例)有相似的推理,包括:属性——实例关系,属性包含,属性体系结构和属性可满足性等。 [0076] (e) based inference attribute: attribute class (Example) similar reasoning, comprising: Properties - Examples of relationships, attributes comprising, architecture and properties and the like properties can be satisfied.

[0077] 五、创建本体的实例 [0077] Fifth, create an instance of body

[0078] 类的结束和个体的开始,决定了最低描述粒度,描述粒度反过来又取决于本体的应用。 Start [0078] of the end of class and individual, determines the minimum granularity of description, description of size, in turn, depends on the application of the body. 所以定义本体实例对下一步本体的应用有直接的关系。 So the definition of ontology instances are directly related to the application of the next body. 在本体中创建一个实例,仅需要声明它是某个类的成员即可。 Create an instance in the body, only you need to declare it to be a member of a class. 如: Such as:

[0079] 〈/owl :Thing rdf :ID =” 香格里拉”〉 [0079] </ owl: Thing rdf: ID = "La">

[0080] <rdf :type rdf :resource =,,# 风景名胜,,/> [0080] <rdf: type rdf: resource = ,, # ,, Sightseeing />

[0081] <rdf :type rdf :resource =” 地理位置”/> [0081] <rdf: type rdf: resource = "Location" />

[0082] 〈/owl:Ting> [0082] </ owl: Ting>

[0083] 声明了个体“香格里拉”,它是“风景名胜”类和“地理位置”类的实例。 [0083] declared individual "Shangri-La", it is an instance of "scenic" class and the "Location" category. 其中,rdf :type出现多次,说明该个体是多个类的实例。 Wherein, rdf: type occur several times, indicating that the subject is a multiple instances of the class.

[0084] 本发明方法云南旅游领域进行了实验验证,首先,以旅游概念作为该旅游本体的Top-level,再粗粒度的将其分为特色食物、住宿、交通方式、风景名胜、地理位置、特产、娱乐活动、少数民族、民族风情等19类,采用人工定义的19个领域概念、34个领域属性和23个领域关系作为本体知识库的构建资源,构成了旅游本体的Middle-level,这些针对每一要素再继续划分,就产生了bottom-level。 [0084] The method of the present invention is the field of tourism in Yunnan conducted experiments, first of all, the concept of tourism as a Top-level tour of the body, and then will be divided into coarse-grained features food, accommodation, transportation, attractions, location, specialty, entertainment, ethnic minorities, ethnic customs and other 19 categories, using 19 fields of artificial concepts defined 34 areas and 23 areas of property relations as a resource to build the body of knowledge, constitutes a Middle-level tourism body, these for each element and then continue to divide, they produce a bottom-level. 然后对知识库进行实例的扩充,构建旅游本体知识库。 Then expand the knowledge base instance, to build tourism ontologies.

[0085] 针对提出的以上方法在云南旅游领域进行了实验验证,具体步骤如下: [0085] carried out for the above method proposed in the field of tourism Yunnan experiments, the following steps:

[0086] 步骤al、人工收集了云南旅游的常用问题1346条。 [0086] Step al, artificial collection of common problems of tourism in Yunnan 1346.

[0087] 步骤a2、对问句进行预处理,主要是将步骤al的问题进行分类,将问题分为:景点、小吃、酒店、风土民情、交通、导购23类。 [0087] Step a2, pretreatment of questions, the main problem is to classify step al, the problem is divided into: attractions, snacks, hotels, local customs, transportation, shopping guide 23 categories.

[0088] 步骤a3、人工自定4条语义规则,用以进行规则推理。 [0088] Step a3, artificial custom four semantic rules, the rules for reasoning. 其中,K(x, y)表示x是y的上位概念,S(x, y)表示X和y是相似概念,Α(χ, ζ)表示ζ是χ的属性。 Wherein, K (x, y) represents the x y is the superordinate concept, S (x, y) represents the X and y are similar concepts, Α (χ, ζ) represents [zeta] [chi] is the attribute. x(R) =Z表示概念χ和概念ζ有角色关系R。 x (R) = Z schematic χ ζ and concepts have role relation R.

[0089] 自定义规则如下: [0089] Custom rules are as follows:

[0090] 【Rule I :K(x, y), A(χ, ζ) — A(y, ζ)】 [0090] [Rule I: K (x, y), A (χ, ζ) - A (y, ζ)]

[0091] 代表上位概念有的属性,下位概念也有该属性。 [0091] Representative properties of some superordinate concept, a subordinate concept also has this property. 例如Κ(动物,人),Α(动物,性别)一A(人,性别)。 E.g. K0 (animal, human), [alpha] (animals, sex) a A (person, gender). 动物有性别的属性,动物的下位概念人也有性别的属性。 Animals have sex attributes, subordinate concept human animals also have sex attributes.

[0092] 【Rule 2 :S(x, y), A(χ, ζ) — A(y, ζ)】 [0092] [Rule 2: S (x, y), A (χ, ζ) - A (y, ζ)]

[0093] 代表某概念有的属性,其相似概念也有该属性。 [0093] Some attributes representative of a concept, which has the properties similar concepts.

[0094] 【Rule 3 :K(x, y),x(R) = ζ — y(R) =ζ】 [0094] [Rule 3: K (x, y), x (R) = ζ - y (R) = ζ]

[0095] 代表上位概念χ和概念ζ有角色关系R,则其下位概念I和概念ζ也有角色关系R0 [0095] represents a generic term and concept of ζ have role χ relation R, the I and the subordinate concept which concept also ζ role relation R0

[0096] 【Rule 4 :S(x, y), x(R) = ζ — y (R) = ζ】 [0096] [Rule 4: S (x, y), x (R) = ζ - y (R) = ζ]

[0097] 代表某概念χ和概念ζ有角色关系R,则其相似概念y和概念ζ也有角色关系R。 [0097] Representative of a concept and the concept of ζ have role χ relation R, the concept of which is similar to the concept and y have the role relationship ζ R.

[0098] 步骤a4、对用户问句进行分析,我们可以发现用户的问句本质上主要有如下三种情况: [0098] Step a4, the analysis of user questions, we can find mainly has the following three conditions on the nature of the user's questions:

[0099] (I)主题+实体+属性 [0099] (I) + entity attributes relating +

[0100] (2)主题+事件+角色 [0100] (2) the role of the theme + Event +

[0101] (3)问句是一组无序的关键字 [0101] (3) questions is an unordered set of keywords

[0102] 步骤a5、对于步骤a4中的(I),在进行答案抽取的时候可能出现以下三种情况: [0102] Step a5, a4 to the step (I), the answer extraction is performed when the following three conditions may occur:

[0103] a.领域知识owl文档中有其对应的具体节点: . [0103] a knowledge owl art document specific corresponding node:

[0104] b.没有具体对应得节点,但有明确的领域主体概念:[0105] c.没有具体对应得节点,但可以找到相似主体概念: . [0104] b have no specific correspondence node, but there is a clear main concept art: [0105] c is not particularly give the corresponding node, but can be found similar to the concept of the body:

[0106] 对于a,直接从owl文档中提取该节点的一段文本、对于b,可以利用步骤a3中的RULE I,采用如下推理算法; [0106] For a, the node extracts a text document directly from the owl, to B, may be utilized in step a3 RULE I, using the following reasoning algorithm;

[0107] 设主体概念为C,属性为A在这里我们把“实体+属性”和“事件+角色”统称为属性。 [0107] set the main concept is C, A and here we attribute to the "Entity + Properties" and "Event + role" collectively referred to as attributes.

[0108] stepl Ψ — upper (C) ; ( Φ 的上位节点赋给Ψ) [0108] stepl Ψ - upper (C); ([Psi] is assigned to an upper node Φ)

[0109] step2 If Ψ具有属性A,则owl中有该属性节,找到答案。 [0109] step2 If Ψ has property A, the owl has the attributes section, find the answer.

[0110] else, if Ψ不是根节点,贝丨JC — Ψ ,转到stepl。 [0110] else, if Ψ is not the root, Tony Shu JC - Ψ, go to stepl.

[0111] else,没有答案,记录本次回答,提交管理员维护。 [0111] else, there is no answer, the records of this answer, submit an administrative maintenance.

[0112] 对于C,可以利用步骤a3中的Rule 2,采用如下推理算法: [0112] For C, step a3 may be utilized in Rule 2, inference algorithm follows:

[0113] 设主体概念为C,其相似主体概念S,属性为A。 [0113] provided the concept of the body C, which is similar to the concept of the body S, the attribute is A.

[0114] stepl找到其相似主体概念S, [0114] stepl found which is similar to the concept of the body S,

[0115] step2 if S具有属性A,则owl中有该属性节,找到答案。 [0115] step2 if S has attribute A, the owl has the attribute section, to find the answer.

[0116] else if S 不是根节点,贝丨J Φ —upper ⑶,(S) — Φ,转到step2。 [0116] else if S is not the root, shell Shu J Φ -upper ⑶, (S) - Φ, go to step2.

[0117] else没有答案,记录本次问答,提交管理员维护 [0117] else there is no answer, the records of this Q & A, filed administrators maintain

[0118] 步骤a6、对于步骤a4中的(2),在进行答案抽取的时候可能出现以下三种情况: [0118] Step a6, in step a4 for (2), the answer extraction is performed when the following three conditions may occur:

[0119] 对于a,直接从owl文档中提取该节点的一段文本。 [0119] For a, the node extracts a text from the document owl. 对于b,可以利用步骤a3中的Rule 3找其先辈节点有否对应的事件和角色。 For b, you can use the steps in Rule 3 a3 find its ancestor node whether the corresponding event and roles. 对于C,可利用步骤a3中的rule4先找主体概念Φ在本体库中的相似概念Ψ,然后转为情况a、b之一进行处理。 For C, can use in step a3 rule4 first find the concept of the body in the body of the library Φ similar concepts Ψ, then into the case a, b, one for processing. 具体算法采取步骤a5中的算法,其中Rule I改为Rule 2,Rule 3改为Rule 4即可。 The algorithm in the algorithm to take steps a5, wherein Rule I to Rule 2, Rule 3 Rule 4 can be changed.

[0120] 步骤a7、对于步骤a4中的(3)问句是一组无序的关键字的时候,在答案抽取时采用进一步与用户交互的策略或采用基于信息检索的方式进行抽取。 [0120] Step A7, for the step a4 (3) a set of keywords question is disordered when the strategy was adopted to further interact with the user when using the answer extraction or decimating mode based information retrieval.

[0121] 实验结果如表I所示。 [0121] The results are shown in Table I below.

[0122] 表I基于本体推理的旅游问答系统答案抽取方法的实验结果比较 [0122] Table I compares answer extraction method based on the results of the forum ontology reasoning system

[0123] [0123]

Figure CN101436206BD00091

[0124] 从实验结果可以看出,基于本体推理的旅游问系统答案抽取的准确率达到了81. 35%,而召回率达到了90. 19%在测试过程中我们发现,知识库中的相关知识的详细程度以及问句分析的准确度都将直接对答案抽取的准确性。 [0124] As can be seen from the results, based on the accuracy of ontology reasoning tourism ask answer extraction system reached 81.35%, while the recall rate reached 90.19 percent in the testing process, we found that in a knowledge base the level of detail and knowledge of the accuracy of the analysis will direct questions to answer extraction accuracy.

Claims (1)

1. 一种基于本体推理的旅游问答系统答案抽取方法,其特征在于第一步以旅游概念作为该旅游本体的顶层,以领域概念、领域属性和领域关系作为本体知识库的构建资源,构成旅游本体的中层,针对每一要素再继续划分,产生底层;然后对知识库进行实例的扩充,构建旅游本体知识库,并对用户问句进行分析;第二步将基于语义规则推理与信息检索相结合进行答案抽取;最后再根据不同的问句类型设计相应的答案抽取算法; 该方法具有包括有: (1)人工定义旅游领域中的概念、属性和关系,并构建旅游领域本体知识库,最后再对本体的一致性进行检验: (2)利用步骤(I)的本体知识库中的语义信息对用户问句进行语义消歧; (3)人工自定义旅游领域中的语义规则;自定义语义规则如下: [Rule I :K(x, y), A(x, z)- > A(y, z)] 代表上位概念有的属性,下位概念 CLAIMS 1. A method for answer extraction system Forum based ontology reasoning, characterized in that the first step in the concept of travel as the top of the travel body, to the concept of the art, the relationship between the field of the art as building properties and resource ontology knowledge base, configured tourism the middle of the body, for each element and then continue to divide, produce the bottom; then expand the knowledge base for instance, to build tourism ontologies, and analyzes user questions; the second step will be based on the rules of inference and semantic information retrieval phase performed in conjunction with answer extraction; finally design the appropriate answer extraction algorithms depending on the type of question; the method comprising: (1) defined in the field of tourism artificial concepts, relations and attributes, and build tourism domain ontology knowledge, and finally and then tested for consistency of body: (2) user questions semantic disambiguation ontologies semantic information in step (I) in; (3) custom artificial semantic rule in the field of tourism; custom semantics rules are as follows: [rule I: K (x, y), A (x, z) -> A (y, z)] some attributes representative of a broader concept, a subordinate concept 有该属性; [Rule 2 :S(x, y), A(x, z)- > A(y, z)] 代表某个概念有的属性,其相似概念也有该属性: [Rule 3 :K(X,y),X(R) = z- > y (R) = z] 代表上位概念x和概念z有角色关系R,则其下位概念I和概念z也有角色关系R ; [Rule 4 :S(X,y),X(R) = z- > y (R) = z] 代表某个概念x和概念z有角色关系R,则其相似概念I和概念z也有角色关系R ;其中上述的K(x, y)表示X是y的上位概念,S(x, y)表示x和y是相似概念,A(x, z)表示z是X的属性,X (R) =Z表示概念X和概念z有角色关系R : (4)基于步骤(2)的问句分析结果,采用基于步骤(3)中的语义规则的推理和信息检索相结合的方法在步骤(I)的本体知识库中抽取答案: (5)根据步骤(2)中的不同的问句类型,设计相应的答案抽取算法,在不降低答案抽取速度的基础上,提高系统的响应率和召回率。 This property has; [Rule 2: S (x, y), A (x, z) -> A (y, z)] representative of a concept and some properties, which also have the property similar concepts: [Rule 3: K (x, y), x (R) = z-> y (R) = z] Representative generic term x and concepts z has a role relationships R, then its lower concept I and concepts z have the role relationship R; [Rule 4: S (x, y), x (R) = z-> y (R) = z] x represent some concepts and concept relationships z has a role R, which is similar to the concept and the concept I z have the relationship role R & lt; wherein the above of K (x, y) represents the X is a superordinate concept of y, S (x, y) represents the x and y are similar concepts, a (x, z) represents a z is the property of X, X (R) = Z represents a concept X and z have the concept of character relation R: (4) based on the step (2) analysis of the question, using the method based reasoning steps and information retrieval semantic rules (3) of the combined body of knowledge in the step (I), answer extraction library: (5) according to step a different question types (2), the corresponding answer extraction algorithm design, without reducing the speed of answer extraction, to improve the response and recall system.
CN 200810233734 2008-12-22 2008-12-22 Tourism request-answer system answer abstracting method based on ontology reasoning CN101436206B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 200810233734 CN101436206B (en) 2008-12-22 2008-12-22 Tourism request-answer system answer abstracting method based on ontology reasoning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 200810233734 CN101436206B (en) 2008-12-22 2008-12-22 Tourism request-answer system answer abstracting method based on ontology reasoning

Publications (2)

Publication Number Publication Date
CN101436206A CN101436206A (en) 2009-05-20
CN101436206B true CN101436206B (en) 2012-09-05

Family

ID=40710644

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 200810233734 CN101436206B (en) 2008-12-22 2008-12-22 Tourism request-answer system answer abstracting method based on ontology reasoning

Country Status (1)

Country Link
CN (1) CN101436206B (en)

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8849828B2 (en) * 2011-09-30 2014-09-30 International Business Machines Corporation Refinement and calibration mechanism for improving classification of information assets
CN102637192A (en) * 2012-02-17 2012-08-15 清华大学 Method for answering with natural language
CN103294770A (en) * 2013-05-06 2013-09-11 清华大学 Human-object interaction method based on tag recognition and natural language semantic analysis
CN104216913B (en) * 2013-06-04 2019-01-04 Sap欧洲公司 Question answering method, system and computer-readable medium
CN103593335A (en) * 2013-09-05 2014-02-19 姜赢 Chinese semantic proofreading method based on ontology consistency verification and reasoning
CN104504023B (en) * 2014-12-12 2017-08-04 广西师范大学 A subjective question computer automatic marking method based high precision of domain ontology
CN104636465B (en) * 2015-02-10 2018-11-16 百度在线网络技术(北京)有限公司 Snippet generation method, apparatus and corresponding display methods
CN105354180B (en) * 2015-08-26 2019-01-04 欧阳江 A kind of method and system for realizing open Semantic interaction service
CN105528437B (en) * 2015-12-17 2018-11-23 浙江大学 Yourself method of constructing a system configuration based on the extracted text information
CN106250366B (en) * 2016-07-21 2019-04-19 北京光年无限科技有限公司 A kind of data processing method and system for question answering system
CN106909662A (en) * 2017-02-27 2017-06-30 腾讯科技(上海)有限公司 Knowledge graph construction method and device

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
余正涛 等.受限域FAQ中文问答系统研究.计算机研究与发展.2007,388-393.
余正涛 等.基于模式学习的中文问答系统答案抽取方法.吉林大学学报(工学版)38 1.2008,38(1),142-147.
陈康 等.受限领域问答系统的中文问句分析研究.计算机工程34 10.2008,34(10),25-27.

Also Published As

Publication number Publication date
CN101436206A (en) 2009-05-20

Similar Documents

Publication Publication Date Title
Sheth et al. Semantics for the semantic web: The implicit, the formal and the powerful
Kashyap et al. Semantic heterogeneity in global information systems: The role of metadata, context and ontologies
Otero-Cerdeira et al. Ontology matching: A literature review
Sheth et al. Semantic (Web) technology in action: Ontology driven information systems for search, integration, and analysis
Cambria et al. Jumping NLP curves: A review of natural language processing research
He Knowledge discovery through co-word analysis
Formica Concept similarity in formal concept analysis: An information content approach
Luo et al. Building association link network for semantic link on web resources
Wimalasuriya et al. Ontology-based information extraction: An introduction and a survey of current approaches
Zhuge Interactive semantics
Nickel et al. A review of relational machine learning for knowledge graphs
Liu et al. Commonsense reasoning in and over natural language
Martinez-Cruz et al. Ontologies versus relational databases: are they so different? A comparison
US8620890B2 (en) System and method of semantic based searching
Ballatore et al. Geographic knowledge extraction and semantic similarity in OpenStreetMap
Marrese-Taylor et al. A novel deterministic approach for aspect-based opinion mining in tourism products reviews
US7428517B2 (en) Data integration and knowledge management solution
Chen The centrality of pivotal points in the evolution of scientific networks
Nebot et al. Multidimensional integrated ontologies: A framework for designing semantic data warehouses
Gagnon Ontology-based integration of data sources
Fu et al. Semantic imitation in social tagging
Lopez et al. Cross ontology query answering on the semantic web: an initial evaluation
Cheng et al. Relin: relatedness and informativeness-based centrality for entity summarization
Hliaoutakis Semantic similarity measures in MeSH ontology and their application to information retrieval on Medline
US9336306B2 (en) Automatic evaluation and improvement of ontologies for natural language processing tasks

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
C14 Grant of patent or utility model
EXPY Termination of patent right or utility model