没有合适的资源?快使用搜索试试~ 我知道了~
知识图谱介绍1
需积分: 0 0 下载量 58 浏览量
2022-08-04
16:10:06
上传
评论
收藏 280KB PDF 举报
温馨提示
试读
15页
知识图谱介绍1
资源推荐
资源详情
资源评论
Chapter 1
Enterprise Knowledge Graph:
An Introduction
Jose Manuel Gomez-Perez, Jeff Z. Pan, Guido Vetere and Honghan Wu
A knowledge graph consists of a set of interconnected typed
entities and their attributes .
Compared to other knowledge-oriented information systems, the distinctive fea-
tures of knowledge graphs lie in their special combination of knowledge represen-
tation structures, information management processes and search algorithms. The
term ‘Knowledge Graph’ became well known in 2012 when Google started to use
knowledge graph in their search engine, allowing users to search for things, people
or places, rather than just matching strings in the search queries with those in Web
documents. Inspired by the success story of Google, knowledge graphs are gaining
momentum in the world’s leading information companies.
The idea of a knowledge graph is not completely new though. The original idea
dates back to the knowledge representation technique called the Semantic Network.
Later on, researchers in Knowledge Representation and Reasoning (KR) addressed
J.M. Gomez-Perez (
B
)
Expert System, Prof. Waksman 10, 28036 Madrid, Spain
e-mail: jmgomez@expertsystem.com
J.Z. Pan
University of Aberdeen, King’s College, Aberdeen AB24 3UE, UK
e-mail: jeff.z.pan@abdn.ac.uk
G. Vetere
IBM Italia, via Sciangai 53, 00144 Rome, Italy
e-mail: gvetere@it.ibm.com
H. Wu
King’s College London, De Crespigny Park, London SE5 8AF, UK
e-mail: honghan.wu@kcl.ac.uk
© Springer International Publishing Switzerland 2017
J.Z. Pan et al. (eds.), Exploiting Linked Data and Knowledge
Graphs in Large Organizations, DOI 10.1007/978-3-319-45654-6_1
1
2 J.M. Gomez-Perez et al.
some well-known issues on the Semantic Network when standardising the modern
version of Semantic Network, or RDF (Resource Description Frameworks). It turns
out that knowledge representation techniques, such as Knowledge Graph or Semantic
Network, are useful not only for Web search, but also in many other systems and
applications, including enterprise information management. The focus of the book,
therefore, is about constructing, understanding and exploiting knowledge graphs in
large organisations.
The basic unit of a knowledge graph is (the representation of) a singular entity,
such as a football match you are watching, a city you will visit soon or anything you
would like to describe. Each entity might have various attributes. For example, the
attributes of a person include name, birthdate, nationality, etc. Furthermore, entities
are connected to each other by relations; e.g. you follow one of your colleagues
in Twitter. Relations can be used to bridge two separate knowledge graphs. For
example, by saying that your Twitter ID and the ID on your driving license are
denoting one and the same person, this actually interlinks Twitter data with the
information space in the driver licensing agency of your country. Not surprisingly,
each entity needs an identification to distinguish one another. This is the final jigsaw
in the knowledge representation of knowledge graphs. Note that to facilitate the
interlinking between various knowledge graphs, the entity IDs need to be globally
unique. Types of entities and relations are defined in some machine-understandable
dictionaries called ontologies. The standard ontology language is called OWL (Web
Ontology Language).
The quality of a knowledge graph is crucial for its applications. For example, a
knowledge graph should be consistent. In the above example, it could be the case
that your contact address in your driving license is different than that in your Twitter
profile. To create a knowledge graph connecting these two information spaces, such
inconsistency should be resolved by keeping the correct one. In addition to consis-
tency, one also needs to consider correctness, and coverage of knowledge graphs,
as well as efficiency, fault tolerance and scalability of services based on knowledge
graphs. Many of those aspects are related to, among others, the schema (ontology)
of a knowledge graph.
A knowledge graph has an ontology as its schema defining
the vocabulary used in the knowledge graph .
1.1 A Brief History of Knowledge Graph
1.1.1 The Arrival of Semantic Networks
Knowledge management in early human history was largely shaped by oral com-
munication before the invention of languages, which then allowed human knowl-
1 Enterprise Knowledge Graph: An Introduction 3
edge to be recorded and passed on through generations. One of the first computer-
based knowledge representation approaches are Semantic Networks, which represent
knowledge in the form of interconnected nodes and arcs, where nodes represent
objects, concepts or situations, and edges represent the relations between them,
including is-a (e.g. “a chair is a type of furniture”) and part-of (e.g. “a seat is part of
a chair”).
As regards the origin of Semantic Networks [38], some researchers argue that
Semantic Networks have come from Charles S. Peirce’s existential graphs, while
many of them pay tribute to Quillian, who was the first to introduce Semantic Net-
works in his semantic memory models [194]. Semantic memory refers to general
knowledge (facts, concepts and relationship), such as a chair. It is different from
another kind of long-term memory, i.e. episodic memory, which relates to some
specific events, such as moving a chair. After Quillian, many variants of Semantic
Networks were proposed.
Compared to formal knowledge representation and reasoning formalisms, such
as predicate logics, Semantic Networks are relatively easy to use and maintain. On
the other hand, they suffer from some limitations. For example, there is no formal
syntax and semantics for Quillian’s Semantic Network. This leaves room for users
to have their own interpretations of constructors in Semantic Networks, such as the
is-a relation. This approach may be seen as flexible for some, but it is also criticised
for making it hard to integrate Semantic Networks while preserving their original
meaning. Furthermore, Semantic Networks do not allow users to define the meaning
of labels on nodes and arcs.
1.1.2 From Semantic Networks to Linked Data
RDF (Resource Description Framework) is a modern standard from W3C, addressing
some of the issues related to classic Semantic Networks in terms of the lack of
formal syntax and semantics. For example, the is-a relation can be represented by
the subClassOf property in RDF, the semantics of which is clearly defined in the RDF
specifications. It should be pointed out that RDF does not address all the limitations
of a Semantic Network, e.g. RDF does not allow users to define concepts either. This
is, however, addressed by OWL (Web Ontology Language), a W3C standard for
defining vocabularies for RDF graphs. In OWL, the part-of relation is not a built-in
relation like the subClassOf property. Instead, it is a user-defined relation that can be
expressed by using the existential constructor. Description Logics [18, 184] are the
underpinning of the OWL standard in the Semantic Web. More details of RDF and
OWL can be found in Chap. 2.
Based on RDF and OWL, Linked Data is a common framework to publish and
share data across different applications and domains, where RDF provides a graph-
based data model to describe objects. OWL offers a standard way of defining vocab-
ularies for data annotations. In the Linked Data paradigm, RDF graphs can be linked
剩余14页未读,继续阅读
资源评论
西西里的小裁缝
- 粉丝: 25
- 资源: 292
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功