本书全面、客观地综述在云计算、大数据环境下知识工程领域面临的新的科学问题及技术挑战，展示国内外在知识工程领域所取得的最新进展，特别是对国内学者系统性、开创性的工作予以足够的重视并详细论述。本书针对大数据知识获取、表示、发现、开发与服务问题，围绕七大主题对国内外进展进行综述，包括大数据知识工程引论、知识表示、知识发现、知识管理与搜索、知识的智能建模、知识迁移和转换、知识工程交叉领域，这是在中国人工智能学会领导下，由知识工程与分布专委会组织编写的有关知识工程的具有重要学术价值和技术前瞻性的技术发展报告，适合人工智能及相关专业本科生、研究生、高级专业技术人员、政府科技管理人员阅读。

序 言 —21 世纪的知识工程在哪里跨越 知识工程诞生于20 世纪60～70 年代。知识工程的诞生帮助人工智能摆脱了当时社会对它的信任危机，是人工智能中最接近实际、最接近社会应用的分支。然而传统的知识工程经历了一段短暂的辉煌以后，又逐渐显示出它的诸多不足。这并不是因为知识工程这个研究方向缺乏存在价值，而是因为传统的知识工程理论和技术不能适应当时社会的进步。 进入21 世纪以后，知识工程面临一个崭新的形势。第一，人类社会进入了大数据时代。吴信东等指出：现代知识工程与传统知识工程的一个重要不同点就是后者强调从专家获取知识，而前者更关注从大数据中获取知识。由此吴信东等提出了大数据知识工程的概念。第二，知识工程与互联网的结合也成为一个重要特点。承载着海量数据的互联网是当今世界大数据的典型代表，互联网又是知识服务的最大提供者。在很多情况下，互联网比大型图书馆更受知识索求者的青睐，尤其是因为它能够提供最新、最及时的知识服务。以维基百科为代表的公众知识服务网站贡献巨大。第三，知识工程的规模达到新的高度。从1968 年开始的以Feigenbaum 型的专家系统为代表（知识量以K 计算）的中小规模时期，到20 世纪80～90 年代以Lenat 的Cyc 常识知识库和曹存根的CNKI 为代表（知识量以M 计算）的大规模知识库时期，发展到了21 世纪以知识图谱、互联网、搜索引擎、浏览器为代表（承载和处理知识量以B 计算）的超大规模知识工程时期。根据Netcraft 研究所发表的数据，可以算出这20 年来网站数量增长符合摩尔定律，每15 个月翻一倍，目前正在向着高达T 级的极大规模知识工程前进。第四，知识表示有了事实规范。在知识工程初起时期，各路研究和技术人员为知识表示设计了多种多样、百花齐放的表示形式。经过数十年的实践和比较，语义网本体知识表示OWL、RDF 成为事实上的知识表示规范，而且还有坚实的逻辑基础。第五，海量知·VI·识组织出现了新的形式—知识图谱，它不仅关心知识点本身，而且也关心知识点之间的联系。得力于Freebase 的前瞻性工作和谷歌的强力开发，知识图谱在2012 年闪亮登场。受此启发，短短几年间出现了一大批内容丰富的代表性知识图谱，本书第三章也重点介绍了部分国产知识图谱的研究和开发工作。目前这种各行各业都用知识图谱的趋势，能够促进行业知识管理的技术革命。第六，出于海量知识处理的需要，知识工程进入了数字化时代。把文字和符号知识映射（嵌入）到数字向量空间来实施知识计算，是富有想象力的创举。之前由于数据量大而难以完成的知识处理，现在有办法完成了。但是在语义的精确性和可解释性方面要付出一些代价。第七，知识工程和软件工程技术互相渗透。早在20 世纪晚期，专家系统和更一般的基于知识的系统就已经广泛使用软件工程手段开发，如面向对象的知识系统开发方法。反过来，基于知识的软件工程早就成为重要研究方向。2005 年提出的知件技术提倡把知识从软件中，特别是应用软件中分离出来，形成独立的、计算机可操作的、商品化的、符合行业规范的、可以嵌入软件或硬件中使用的知识模块，称为 知件，知件和硬件、软件一起构成信息产业的三大支柱。由此，知件工程、基于知件的软件工程、知件和软件协工程等都是值得深入研究的课题。第八，MapReduce 的兴起是分布式数据处理技术的新突破，使知识工程中高效的大规模数据处理有了盼头。但是正如人们已经发现的，MapReduce 主要适用于数据密集型知识处理任务，并不是对所有领域、所有问题、所有算法都适用，例如对于兼有分析、推理、递归密集型的任务，此算法框架就不一定适用。 第九，云计算的有效性和重要性已经成为国际共识。传统的云计算模式主要是各类云计算中心及其平台。近年国际上提出云计算要深入到用户终端级，称为雾计算。更有人提出云计算要深入个人计算机层次，称为露计算，从而形成一个有中心、有主干、有分支的云、雾、露计算网络。以上一切都引发人们思考：知识工程的下一步应该是什么？对此，我们首先关心的是：大数据对知识工程的发展有何作用？人们对数据感兴趣是因为从中可以获取知识。很自然，如果是大数据，则对所获取的知识就有了不一样的期待。它有可能是大知识吗？2010 年以后，早在有人VII·提出大数据3V、4V、5V 特征的同时，“大知识”一词就在网上、文章中反复出现，大知识的正面研究空缺了一段时间，最近两年才开始有论文发表。关于大知识和大知识系统概括出了10 个大规模性质，其中“最浓缩”的说法是：“大知识是体积海量、结构复杂、组织精密、可信度高、使用方便、与时俱进、能够解决人类一大类问题的系统性知识。”不妨把它与吴信东对大数据特性的HACE 定理（异构、自治、复杂、演化）做一下对比。不难看出，HACE强调大数据的获取容易而处理、使用难，大知识则与此相反，获取（构造）难而（一旦构造好）使用顺手。本书第二章集中介绍了大知识和大知识工程的有关概念和主要特征。我们相信，学术界和企业界都会对大知识的研究和实践产生越来越大的兴趣。第二个展望涉及新一代知识工程的技术、算法和工具，其中最令人关注的是知识获取自动化。面对海量涌现的信息和知识,浏览器技术需要更新换代，改变每逢查询便推出一大堆网页的做法，代之以能够自动为用户按需从大批网页中实时提炼和综合知识，包括综合利用最新技术来提供解决用户实际问题的方案。进一步，按照我们的设想，未来的浏览器应该类似现在生活中的手机，在网上无所不能。第三，语义网需要来一场革命，因为Bernes Lee 的理想并未完全实现，叫好不叫座。关键是缺少一种以OWL 和RDF 为一边，自然语言为另一边的自动互译技术，从而用起来 很不方便，无法推广。目前有类自然语言和受控自然语言两种技术，但都没有达到能实用于上述自动互译的地步。第四，把网上知识服务进行到底。四千多万篇维基百科文章目前还类似一个个自生自灭的知识孤岛，虽然读者可以通过引文链接穿行于孤岛之间，却无法让孤岛们自发组织起来解决一个个复杂的问题。下海（网下知识自动转换成维基百科文章）和登陆（网上知识的自动归纳、升华）甚至还没有被提上日程。第五，知识图谱技术虽然受到学术界和业界的普遍重视，却仍然处在初级阶段。知识图谱的开发者们目前最关心的是扩张领土，包容知识以多多益善为上。而知识图谱内知识的管理及知识图谱对社会和学术界的服务能力却仍然滞后，更不用说它的理论基础了。对知识图谱的研究应该有一个大的发展。第六，应该鼓励知识工程“上山下乡”，推动微小型分布式知识工程的发展。现在国家很重视微小型企业的·VIII·发展，也很重视农村潜力的开发。我们不仅要有作为心脏的大型云计算，作为主干的用户终端级的雾计算，作为分支的个人计算机露计算，还要有作为神经末梢的手机胞计算（cell computing），让计算能力无孔不入，无处不在。第七，关于数据流、控制流和知识流。传统编程主张分开数据流和控制流，使复杂的程序清晰可读。然而在大数据时代，仅仅分开数据流和控制流是不够的。巨量多变的数据使得传统方法的控制粒度变得过小，复杂性问题难以解决。利用数据之间的语义联系，让流动的数据上升为知识流，将能大幅度提高控制的可读性和有效性，我们称之为知识流程序设计。例如在进程代数中引进逻辑程序设计，对于程序的模块化、可读性、可验证性、可复用性均有好处。第八，知件技术和理论应该有突破性的进展，当然绝不是侵占软件和硬件的“领土”，而是进一步渗透进软硬件的“脉络”以内，溶化于软硬件的“血液”之中。近几年，“软件定义网络”已经发展成“软件定义一切”。但软件的“灵魂”是知识。如果知件理论和技术有足够的发展，我们将看到“软件定义一切”的背后实际是“知件定义一切”。第九，在大力推广公众知识教育的同时，也要花力气推进知识的商品化。知识工程创始人Feigenbaum有一句名言：“知识自身应该像食物和石油一样成为可以出售的商品。知识应该成为国家的新财富。”国家知识产权局就是为此而存在的。但是知识的各种商品特性远未充分体现。金属、石油、粮食都有期货和期权，知识为什么不可以有？前者都可以有交易所，知识为什么不可以有？我们认为，在商品经济充分发达的新世纪，应该以实际措施为知识的商品属性正名。知识的获取和应用是人类永恒的需求，所以知识工程也就是人类永恒的课题。相信新世纪知识工程的发展一定会超出我们的预期。就像太阳每天都从地平线上升起，我们每天也可以看到知识工程灿烂的光焰。人们的肉眼难 以区分昨天和今天的阳光，但是却非常有信心迎接明天比今天更加灿烂的知识工程的光焰！ 2020 年8 月 ·IX· 前 言 人类走过了农耕社会、工业社会、信息社会，已经进入智能社会, 进入在动力工具基础上发展智力工具的新阶段。在农耕社会和工业社会,人类的生产主要基于物质和能量的动力工具，并且得到了极大地发展。今天，劳动工具转向了基于数据、信息、知识、价值和智能的智力工具，人口红利和劳动力红利不那么明显了，智能的红利来了。人工智能再一次成为社会各界关注的焦点，甚至要把我们所处的时代用“智能”来命名，这距离人工智能这一概念首次提出已经过去了六十多年。在这期间，我们经历了人工智能发展的两落三起，我们可以历史上知识工程研究的过程，没有历史就没有未来。1994 年美国工程院院士爱德华指出利用自动机对知识进行获取、操作和利用的工程称之为知识工程，1999 年爱德华教授在获得图灵奖时发表演讲，主题为How the What become the How，演讲中再次提出什么叫知识？知识就是力量吗？人工智能的这一分支学科—知识工程也在快速发展之中。在知识工程领域，爱德华教授跟劳伊雷迪教授一起获得了图灵奖，还有来到中国很多次的西蒙教授，西蒙教授不但获得了图灵奖，还获得了诺贝尔奖，还获得了美国心理学会的终生成就奖。我国的吴文俊院士关于几何定理机器的证明也被全世界认可。这里还要提到一个重要人物，维基百科的创始人多纳尔·威尔士把知识通过网络提供给了全社会，维基百科也是了不起的知识工程。 知识工程是智能科学和技术的分支学科，是人工智能时代最有意义的课题之一。知识工程是研究如何用机器代替人，实现知识的表示、获取、推理、决策和应用的工程，是可解释的人工智能，包括机器定理证明、专家系统、机器博弈、数据挖掘和知识发现、不确定性推理、领域知识库，还包括数字图书馆、维基百科、知识图谱、网络教育等大型知识工程。知识工程不仅研究如何获取、表示、组织和存储知识，还研究如何实现知识型工作的自动化，更要研究如何运用知识和创造新知识。所以几年前就有人提出了知识工程的自动化，现在大家发现这个定义已经不行了。知识工程要关注学习，智能植根于教育，要依靠有指导的学习和自主学习，把人类已有的群体智能—知识更好地变成个体智能和机器人智能。知识工程对教育的挑战，最明显的问题是人类教师的工作会不会被机器人替代？我们从小就被教育，知识就是力量，这是培根讲的。费根鲍姆说这句话不太准确，他主张知识蕴含力量。教育工程是知识工程的典型代表，如果说土木工程是培养修路铺桥，建筑工程是培养盖房盖楼，软件工程是在高效平台上生产新的软件，知识工程就是为了教育人或者机器人学知识长本领，所以学习才是最基本的智能。对人而言，学习是理解知识、形成记忆的交互过程，记忆是学习的结果；对无意识的机器人而言，学习的本质是记忆和解释，知识工程应该研究机器人的 学习和记忆。当人工智能和知识工程走进学校的时候，传统的教育就要改革了。学知识就是学文化、学文明，学习人类积累的群体智能，学知识和学外语、学开车、学游泳没有本质的区别，学习的动机比内容更重要，知识不应该是冷的，知识的运用比知识的容量更重要。因此对教育工程来说，不仅是电子书包和网络教学，更重要的是个性化教学，将来很多的教师工作都会被机器人替代，让教育智能机器人进入课堂，教师成为教练，让学生走向成功。知识工程是一个认知交互的工程。如何听、说、看、想表现在大脑里面就是如何思维，如何变成可解释的智能。因此我一再强调研究脑认知应该研究记忆认知、计算认知和交互认知。人类对计算机近乎贪得无厌的需求阻碍了我们对记忆智能的发展，导致计算智能已经超过人类，为什么不去研究记忆智能呢？AlphaGo 程序有很大的功耗，为什么我们不想想记忆智能可以降低它的功耗呢？这就需要知道三个记忆区：感觉记忆、工作记忆和长期记忆是如何协同工作的？我想这也是人工智能时代知识工程最有意义的课题之一，这就是我最近几年为什么乐此不疲的研发驾驶脑的原因。知识工程进入寻常百姓家。当机器人在和你聊天时，它在进行听、说、读、写、问、译等操作，机器人可以给你开药方，机器人也可以为孩子进行心理咨询，做辅导。我想和东南大学合作，让孩子们在手机上玩游戏就可以预测他有没有自闭症或者多动症，家里看到的电视里面的主持人可能是一个机器人，所以知识工程已经在我们的寻常百姓家里了。 创新驱动，智能担当，教育先行。我国《新一代人工智能发展规划》明确了我国人工智能发展三步走的战略发展目标，当形形色色的各种机器人成为人类认知自然与社会、扩展智力，走向智慧生活的重要伴侣的时候，当机器人和智能系统无处不在地改变着人类的生产活动、经济活动和社会生活的时候，中国要在2030 年成为世界主要人工智能创新中心。为此，中国人工智能学会策划并组织编制以人工智能系列白皮书丛书，本书是该系列丛书之一，期望本书的出版能推动我国知识工程领域的科学研究与技术发展，造福智能 社会。 李徳毅 2020 年8 月

目 录 第一部分 大数据知识工程引论 第1 章 传统知识工程到大数据知识工程 ····················································.2 1.1 人工智能与知识工程 ······································································.3 1.2 知识工程的发展 ·············································································.4 1.2.1 实验性系统阶段（1965～1974 年）··············································.4 1.2.2 MYCIN 阶段（1975～1980 年） ··················································.5 1.2.3 知识工程的应用阶段（1980 年至今） ··········································.5 1.3 大数据知识工程概念与处理框架 ···················································.5 1.3.1 HACE 定理 ···············································································.6 1.3.2 大数据知识工程模型—BigKE ···················································.8 1.4 大数据知识工程的应用场景 ·························································.10 1.4.1 大数据知识工程在电子商务领域的应用 ·······································.10 1.4.2 大数据知识工程在教育领域的应用 ·············································.10 1.4.3 大数据知识工程在医学领域的应用 ·············································.11 1.4.4 大数据知识工程在决策领域的应用 ·············································.12 1.4.5 大数据知识工程在华谱系统的应用 ·············································.12 1.5 小结：大人工智能时代 ·································································.13 参考资料 ································································································.13 第2 章 大知识与大知识工程 ·····································································.16 2.1 从大数据到大知识 ········································································.17 2.2 大知识与大知识系统 ·····································································.18 2.3 大知识工程 ···················································································.23 2.4 大知识工程实践和原型系统 ·························································.25 ·XIV· 2.5 小结 ·······························································································.26 参考资料 ································································································.27 第二部分 知 识 表 示 第3 章 开放知识图谱 ················································································.33 3.1 知识图谱简介 ················································································.33 3.2 知识图谱的价值 ············································································.35 3.2.1 辅助搜索 ·················································································.35 3.2.2 辅助问答 ·················································································.36 3.2.3 辅助大数据分析 ·······································································.37 3.2.4 辅助语言理解 ··········································································.37 3.2.5 辅助设备互联 ··········································································.38 3.3 开放的知识图谱项目 ·····································································.40 3.3.1 早期的知识库项目 ····································································.40 3.3.2 互联网时代的知识图谱 ·····························································.41 3.4 中文领域开放知识图谱：OpenKG ················································.44 3.4.1 中文领域知识图谱的开放现状 ····················································.44 3.4.2 开放的中文领域知识图谱Schema ···············································.45 3.4.3 中文开放知识图谱众包平台 ·······················································.46 3.5 多源信息融合与开放网络知识计算 ··············································.46 3.5.1 开放知识网络构建 ····································································.47 3.5.2 开放知识融合 ··········································································.48 3.5.3 开放知识网络特征 ····································································.49 3.6 大规模复杂网络的多元结构知识发现 ···········································.53 3.6.1 SBM 研究进展 ·········································································.56 3.6.2 小结 ·······················································································.58 参考资料 ································································································.59 第4 章 非规范知识表示与处理 ·································································.62 ·XV· 4.1 非规范知识处理的基本理论概述 ··················································.64 4.2 非规范知识处理研究成果概述 ······················································.66 4.3 基于模糊性的知识表示与学习 ······················································.71 4.3.1 知识表示参数的神经网络净化 ····················································.71 4.3.2 模糊极值熵理论 ·······································································.71 4.3.3 无监督学习的0.5-偏离模型和相应的计算策略 ······························.72 4.4 基于模糊粗糙集的建模与学习 ······················································.73 4.4.1 知识约减 ·················································································.73 4.4.2 模糊集合上下近似算子的构造 ····················································.73 4.4.3 基于模糊粗糙集的增量属性约简 ·················································.74 4.4.4 核函数框架下的模糊粗糙集 ·······················································.74 4.5 粗糙近似算子的构建与多粒度多标记模型 ···································.74 4.5.1 多粒度标记粗糙集知识表示模型 ·················································.74 4.5.2 粗糙集理论和证据理论 ·····························································.75 4.5.3 粗糙近似算子公理化 ·································································.75 4.5.4 多粒度模式知识发现的数据建模方法 ··········································.76 4.6 复杂环境下信息系统知识不确定度量 ···········································.76 4.6.1 不协调序信息系统的最大分布约简 ·············································.76 4.6.2 不确定性知识度量 ····································································.77 4.7 不完备与不一致信息的概念学习 ··················································.78 4.7.1 不完备概念知识表示 ·································································.78 4.7.2 不一致信息的概念知识表示 ·······················································.78 4.7.3 概念认知学习 ··········································································.78 参考资料 ································································································.79 第5 章 因素空间与知识表示的数学理论 ···················································.82 5.1 发展智能科学需要智能数学 ·························································.82 5.2 因素空间的内容、意义与方法 ······················································.85 ·XVI· 5.3 因素空间对知识工程的基本构想 ··················································.94 参考资料 ································································································.98 第6 章 知识粒计算—理论、模型与方法 ·············································.100 6.1 数据粒化 ·····················································································.101 6.2 粒计算模型 ·················································································.101 6.2.1 模糊集模型 ············································································.101 6.2.2 粗糙集模型 ············································································.102 6.2.3 商空间模型 ············································································.102 6.2.4 云模型 ··················································································.102 6.2.5 三支决策模型 ········································································.103 6.3 不确定性度量 ··············································································.103 6.4 粒计算推理 ·················································································.103 6.5 多粒度模型与方法 ······································································.104 6.5.1 多粒度计算模型 ·····································································.104 6.5.2 多粒度不确定性度量 ·······························································.104 6.5.3 多粒度模式发现与融合 ···························································.105 6.6 粒计算的应用研究 ······································································.105 参考资料 ······························································································.106 第7 章 时空知识表示与推理 ···································································.108 7.1 定性空间关系模型 ······································································.109 7.2 定性空间关系推理 ······································································.109 7.3 时空知识推理应用 ······································································.110 参考资料 ······························································································.111 第三部分 知 识 发 现 第8 章 大数据知识发现—挑战与应对 ·················································.115 8.1 大数据知识发现技术挑战 ···························································.115 ·XVII· 8.1.1 描述与存储的挑战 ··································································.115 8.1.2 分析与理解的挑战 ··································································.116 8.1.3 挖掘与预测的挑战 ··································································.116 8.2 大数据知识发现技术研究成果 ····················································.116 8.2.1 大数据处理技术 ·····································································.116 8.2.2 大数据挖掘与知识发现 ···························································.119 8.3 大数据实践 ·················································································.121 8.4 小结 ·····························································································.122 参考资料 ······························································································.123 第9 章 大数据知识发现—理论与技术 ·················································.125 9.1 描述性统计方法 ··········································································.125 9.2 可视化数据挖掘方法 ···································································.126 9.3 机器学习方法 ··············································································.127 9.4 国内研究现状 ··············································································.129 参考资料 ······························································································.130 第10 章 富格式文本中的知识发现 ··························································.135 10.1 背景 ···························································································.135 10.2 文档结构识别 ············································································.140 10.3 自然语言语义提取 ····································································.142 10.4 表格语义提取 ············································································.146 参考资料 ······························································································.147 第四部分 知识管理与搜索 第11 章 大数据挖掘与知识管理 ······························································.154 11.1 研究背景 ···················································································.154 11.2 智能知识管理研究概述 ·····························································.156 11.3 智能知识管理基本概念 ·····························································.157 ·XVIII· 11.4 智能知识管理研究现状 ·····························································.160 11.5 未来研究方向 ············································································.162 参考资料 ······························································································.163 第12 章 智能知识管理·············································································.164 12.1 引言 ···························································································.164 12.2 智能知识管理研究概况 ·····························································.166 12.3 智能知识的挖掘算法与技术 ······················································.167 12.3.1 转化规则挖掘的方法 ·····························································.167 12.3.2 基于多目标线性规划的二次挖掘方法及其在信用卡客户流失 管理中的应用 ·······································································.168 12.3.3 智能知识管理系统设计技术 ···················································.168 12.4 知识可拓优化技术 ····································································.169 12.4.1 研究概况 ·············································································.169 12.4.2 知识可拓自分类优化技术 ·······················································.169 12.4.3 知识可拓自聚类优化技术 ·······················································.171 12.4.4 知识可拓自识别优化技术 ·······················································.171 12.5 小结 ···························································································.172 参考资料 ······························································································.175 第13 章 基于认知的多媒体大数据驱动知识搜索 ····································.178 13.1 大数据驱动的知识搜索 ·····························································.178 13.2 大数据异质媒体搜索环境下的认知行为规律分析 ····················.180 13.2.1 传统桌面搜索场景下的用户认知行为 ······································.181 13.2.2 泛在设备搜索场景下的用户认知行为 ······································.182 13.3 面向大数据泛在搜索环境的异质媒体搜索 ·······························.183 13.3.1 面向异质媒体资源的表示学习方法 ··········································.183 13.3.2 图像资源表示学习 ································································.184 13.3.3 音频资源表示学习 ································································.184 ·XIX· 13.3.4 视频资源表示学习 ································································.184 13.3.5 面向带噪声数据的学习排序方法 ·············································.185 13.3.6 数据驱动的带噪声数据学习排序 ·············································.185 13.3.7 算法驱动的带噪声数据学习排序 ·············································.185 13.4 面向复杂查询的异质媒体搜索技术 ··········································.186 13.4.1 复杂查询的搜索意图理解与推理 ·············································.186 13.4.2 基于知识图谱的搜索意图理解 ················································.186 13.4.3 基于智能问答的搜索意图理解 ················································.187 13.4.4 异质媒体数据聚合与深度排序 ················································.188 13.4.5 泛在搜索场景用户交互形式分析建模 ······································.188 第五部分 知识的智能建模 第14 章 智能体系统 ················································································.192 14.1 概述 ···························································································.192 14.2 智能体ABGP 模型 ····································································.194 14.2.1 环境感知 ·············································································.194 14.2.2 信念 ····················································································.196 14.2.3 目标 ····················································································.196 14.2.4 规划 ····················································································.197 14.3 智能体系统结构 ········································································.197 14.4 脑机融合推理与决策 ·································································.198 14.4.1 脑机融合的自动推理 ·····························································.200 14.4.2 脑机融合的协同决策 ·····························································.200 参考资料 ······························································································.201 第15 章 多智能体协同与智能博弈 ··························································.203 15.1 概述 ···························································································.203 15.2 多智能体学习 ············································································.207 ·XX· 15.3 多智能体协作 ············································································.208 15.4 分布式规划 ················································································.209 15.5 算法博弈论 ················································································.210 15.6 安全博弈论 ················································································.211 15.7 智能对弈 ···················································································.212 15.8 未来展望 ···················································································.213 参考资料 ······························································································.213 第16 章 脑机融合的信息处理 ·································································.217 16.1 引言 ···························································································.217 16.2 脑机融合的混合智能 ·································································.218 16.3 混合智能的信息回路模型 ·························································.220 16.4 脑机融合的信息处理框架 ·························································.222 16.5 脑机融合信息处理的实现思路初探 ··········································.224 16.6 小结 ···························································································.226 参考资料 ······························································································.227 第六部分 知识迁移和转换 第17 章 迁移学习 ····················································································.229 17.1 迁移学习算法 ············································································.230 17.1.1 基于样本实例重新加权的迁移学习算法 ···································.230 17.1.2 基于模型融合集成的迁移学习算法 ··········································.230 17.1.3 基于特征选择的迁移算法 ·······················································.231 17.1.4 基于特征映射的迁移算法 ·······················································.231 17.1.5 基于深度学习的迁移学习算法 ················································.232 17.2 迁移学习研究前景及挑战 ·························································.233 参考资料 ······························································································.235 第18 章 可拓知识工程与知识转换 ··························································.240 ·XXI· 18.1 可拓信息—知识—策略形式化体系 ··········································.241 18.2 基于可拓规则的知识表示方法 ··················································.242 18.3 可拓知识获取 ············································································.244 18.3.1 可拓分类知识获取 ································································.245 18.3.2 传导知识获取 ·······································································.246 18.3.3 基于知识库的可拓知识获取 ···················································.247 18.4 不相容问题求解 ········································································.247 18.4.1 可拓策略生成方法 ································································.248 18.4.2 可拓策略生成系统 ································································.250 18.5 小结 ···························································································.251 参考资料 ······························································································.251 第七部分 知识工程交叉领域 第19 章 基于知识的软件工程 ·································································.255 19.1 全过程基于知识方法 ·································································.255 19.2 基于环境建模的需求获取和分析 ··············································.258 19.3 基于深度学习的软件代码分析和生成 ·······································.261 19.4 知件和基于知件的软件工程 ······················································.262 参考资料 ······························································································.264 第20 章 农业大数据知识服务 ·································································.266 20.1 农业大数据的知识获取 ·····························································.267 20.1.1 基于智能引导的人工知识获取 ················································.267 20.1.2 自动和半自动知识获取 ··························································.268 20.2 农业知识库构建 ········································································.269 20.2.1 知识表示策略 ·······································································.269 20.2.2 推理机制 ·············································································.270 20.2.3 专家系统开发平台 ································································.270 ·XXII· 20.2.4 知识发现系统开发平台 ··························································.271 20.3 农业知识服务 ············································································.271 参考资料 ······························································································.275 第21 章 深度学习与自然语言处理 ··························································.278 21.1 自然语言处理 ············································································.278 21.2 深度学习 ···················································································.279 21.2.1 深度学习的提出与发展 ··························································.279 21.2.2 深度学习的基本原理 ·····························································.280 21.3 深度学习在自然语言处理中面临的问题 ···································.281 21.4 深度学习在自然语言处理中的应用 ··········································.282 21.5 未来深度学习在自然语言处理中的发展 ···································.284 21.6 小结 ···························································································.284 参考资料 ······························································································.285 第22 章 互联网谣言检测与知识辨伪 ······················································.287 22.1 技术背景 ···················································································.288 22.2 谣言检测相关定义 ····································································.290 22.3 谣言检测任务的难点 ·································································.291 22.4 谣言检测最新技术 ····································································.292 22.4.1 基于深度学习的谣言文本模式挖掘 ··········································.292 22.4.2 基于深度学习的谣言视觉模式挖掘 ··········································.293 22.5 小结 ···························································································.294 参考资料 ······························································································.294 第八部分 展 望 第23 章 知识工程领域未来展望 ······························································.299 23.1 知识表示与推理 ········································································.300 23.2 知识驱动的智能技术 ·································································.302 ·XXIII· 23.3 机器学习可解释性研究 ·····························································.303 23.4 深度学习与传统方法的结合 ······················································.305 23.5 小样本学习研究 ········································································.307 23.6 小结 ···························································································.309 参考资料 ······························································································.309 后记 ············································································································.313 外国人名中英文对照表 ·······································································.315 外国机构名中英文对照表 ····································································.316