Specifying architecture of knowledge graph with data graph, information graph, knowledge graph and wisdom graph

Full text

Specifying Architecture of Knowledge Graph with
Data Graph, Information Graph, Knowledge Graph
and Wisdom Graph
Yucong Duan1
State Key Laboratory of Marine
Resource Utilization in the South China
Sea, College of Information and
Technology, Hainan University
duanyucong@hotmail.com
Zhangbing Zhou4
China University of Geosciences
(Beijing), Beijing, China
zhangbing.zhou@gmail.com
Lixu Shao2*
State Key Laboratory of Marine
Resource Utilization in the South China
Sea, College of Information and
Technology, Hainan University
751486692@qq.com
Quan Zou5
College of Computer Science
Tianjin University
zouquan@tju.edu.com
Gongzhu Hu3
Department of Computer Science
Center Michigan University
Michigan, USA
hu1g@cmich.edu
Zhaoxin Lin6
School of business
Iowa State University, USA
zxlin@iastate.edu
Abstract—Knowledge graphs have been widely adopted, in large
part owing to their schema-less nature. It enables knowledge
graphs to grow seamlessly and allows for new relationships and
entities as needed. Knowledge graph has become a powerful tool
to represent knowledge in the form of a labelled directed graph
and to give semantics to textual information. A knowledge graph
is a graph constructed by representing each item, entity and user
as nodes, and linking those nodes that interact with each other via
edges. Knowledge graph has abundant natural semantics and can
contain various and more complete information. Its expression
mechanism is close to natural language. However, we still lack a
unified definition and standard expression form of knowledge
graph. We propose to clarify the expression of knowledge graph as
a whole. We clarify the architecture of knowledge graph from data,
information, knowledge, and wisdom aspects respectively. We also
propose to specify knowledge graph in a progressive manner as
four basic forms including data graph, information graph,
knowledge graph and wisdom graph.
Keywords- knowledge graph, data, information, knowledge, wisdom
I. INTRODUCTION
There are different kinds of discrete data in the real world we
live in. The data cannot be used if they exist only in the discrete
form. However, this is not worth worrying as we can simply
make the data meaningful by giving a specific environment. At
this point, the data can get real meaning in this specified
environment we supplied. The data are processed to be useful
and presented to us in the form of information, then we can get
a lot of fragmented expressions. For example, "Li Ming's height
is 177cm" is the information is capable of being used to answer
questions about Li Ming’s height. Obviously, getting
information is just a start. With these fragmented expressions,
that is, the conception "information" we mentioned above, we
can combine multiple information to answer more complex
questions about how to do it. By abstracting and converting the
information and the data in a given context and the application
of data and information [2], “Knowledge” shows up. At this
point, by defining the attributes and classification summary, we
can digging the information from the data, and we can
summarize knowledge from the information. Furthermore,
comprehensive knowledge of the same category can be use of
making favorable judgments, precisely predicting, and smartly
planning. Obviously, the utilization of vested knowledge is
beyond its literal meaning of the category, which is what we say,
"wisdom". In this paper, we propose to specify the architecture
of knowledge graph in four aspects including data graph,
information graph, knowledge graph and wisdom graph as well
as the practical significance of the analysis of the four
conceptions. Fig.1 shows the conversion from data to wisdom.
Data existing as discrete elements have no semantics.
Information is data after procession of conceptual mapping and
relational connection. The user access to information after
filtering the valuable information and internalized those
information into knowledge. When information is adequately
assimilated, it produces knowledge, modifies the individual’s
mental store of information and benefits his development and
that of the society in which he lives.
Data Information Knowledge Wisdom
Conceptual mapping
Relational connection
Appropriate executionInternalized
Figure 1. Conversion process from data to wisdom

In the rest of this paper, we firstly elaborate representations of
data graph, information graph, knowledge graph and wisdom
graph in Section 2, 3, 4 and 5respectively. Then we describe the
progressive relationship among data, information, knowledge
and wisdom in Section 6. The related works are elaborated in
Section 7. And we conclude our work in Section 8.
II. REPRESENTATION OF DATA GRAPH
Data are the symbolic representation of observable properties
of the world. Data are obtained by observing the basic individual
items of numbers or other information, but on their own, without
context, they have no meaning. Storing of data does not change
the data itself, but it has many expression forms. As Fig.2 shows,
data can be organized in many different types of data structures,
including arrays, stacks, list links and so on. The data can be
structured, semi-structured and unstructured, relational or non-
relational. Data structures can store a great deal of data in many
different types, including numbers, strings and even other data
structures. Generally, data can be represented as many discrete
elements originally [8]. Fig. 3 shows a series of original discrete
data points and that we can use array, linked list, tree, graph as
well as the combination of these four structures to represent data
respectively. Original discrete data points have no meaning
without context. For example, the value 120 can be clinical
measurement such as heart rate and it can also indicates the
telephone number of the emergency center. Cutting them from
the specific context of situations, we cannot determine what it
means for sure. We use a collection D {d1, d2, …, di, …, dn} to
represent the data sets where di indicates a discrete element. For
example, if we input a collection of a series of discrete elements
describing risk assessment of software engineering it can be
denoted as D (risk). We cannot understand the specific meaning
of each element without context and the internal relationship of
these elements.
Raw
Data
Data
struture
Being stored
array
Link list
tree
graph
Stack/queue
include
Equal to
Data = Storage(Data)
Figure 2. Storage of data does not change data itself
array
120 blue A sara99 red 70xxx
list
120 blue
70 xxx Asara
99 red
tree graph
Combination of
different expressions
Raw data
=
120
xxx
blue
99
A
sara
red70 CC3
string
blue120
CC3
99
redA
70 xxx
sara
120
blue 99
red
A
xxx
sara
=
red
120 blue 99
redA
xxx
A
xxx
CC3
70 sara
Figure 3. Ilustration of data and data expression
III. RRESENTATION OF INFORMATION GRAPH
Information is data that has been given meaning by way of
relational connection. This "meaning" can be useful, but does
not have to be. Items of information include elements of
information and relations between the elements of information.
The elements of information are displayed as nodes and relations
are displayed as lines on the information graph. Information
embodies the understanding of a relationship of some sort and
the essence of information phenomenon has been characterized
as the occurrence of a communication process that takes place
between the sender and the recipient of the message. The
conceptual mapping of different concepts and relationships is
called concept mapping [3]. Giving two concepts O1, O2 and
their associated sets C1 & R1 and C2 & R2, conceptual mapping
is to identify potential pairs: (c1, c2) or (r1, r2 ), where c1 ∈ C1,
c2 ∈ C2 , r1∈R1 and r2∈R2. In this way, concept c1 and
relationship r1 can each be translated into instance c2 and
instance r2 while preserving their original meaning. We define a
conceptual mapping function F acting on two concepts C1 and
C2. The similarity function is defined in two concepts C1 and
C2, and a value between 0 and 1 is calculated to indicate the
similarity between C1 and C2. The logical representation of the
concept Ci is L (Ci) where the function F is evaluated as a
similarity representation of the logical representation between
C1 and C2. The function S is used to define the logical similarity
evaluation on L (C1) and L (C2).
 
There is a need to apply the conceptual semantics of
professionals from different aspects. Correspondingly, different
semantics can be represented by independent logical statements,
and function F can be defined exactly as follows:
 

where i indicates a kind of features and its logical expression is
defined as Li. w(i) is an application collaborative function that is

applied to i (determined by the application professional) to
measure the importance of each i in evaluating the similarity of
C1 and C2. Fig. 4 illustrates that a series of raw data points can
be converted to information through conceptual mapping.
Relationships between the information obtained through
conceptual mapping are consistent with relationships between
the original concepts. On the information graph, there is a simple
combination relationship between data points. The contextual
relevance of the information is limited, and in different contexts
we can establish different classification and combination rules.
In Fig. 5 we can recognize that there are three kinds of data
including risk factors, combination type of these factors and the
corresponding probability of each combination type through
conceptual mapping. Then we can have a more complete
description of risks that can denoted as D (A, C, P) and store the
description in relational database. A indicates factors that may
lead to risk. C indicates the combination type of these factors and
P indicates the corresponding probability of each combination
type. A= {system type, developing experience, software and
hardware equipment}. C includes eight combination types of the
three factors and P includes the corresponding eight probabilities.
Raw Data
d1
d2
dn
Conceptu al mapping
CPT(A)
A
A3
A2
A1
Havingfeatures
Havingfeatures
Havingfeatures
Infor(A)Raw Data
d1
d2
dn
Conceptual mapp ing
CPT(B)
d1
d2
dn
B
B3
B2
B1
Havingfeatures
Havingfeatures
Havingfeatures
d1
d2
dn
Conceptual m apping
CPT(C)
d1
d2
dn
C
C3
C2
C1
Havingfeatures
Havingfeatures
Havingfeatures
d1
d2
dn
Conceptual m apping
CPT(D)
d1
d2
dn
D
D3
D2
D1
Havingfeatures
Havingfeatures
Havingfeatures
d1
d2
dn
A B C
D
Rel1 Rel2
Rel3 Rel4
Infor(A) Infor( B) Infor(C)
Infor(D)
Rel1 Rel2
Rel3 Rel4
Figure 4. Conceptual mapping from data to information
3 years
Well
equipment
Unix
Unix
50%
10 years 10%
70%
60%
40%
80%
30%
20%
Incomplete
equipment
probability
10 years Well
equipment Windows
10 years Incomplete
equipment Unix
probability
10 years
3 years
3 years
3 years
Incomplet e
equipment Windows
Well
equipment
Well
equipment
Incomplete
equipment
Windows
Unix
Windows
Hardware and
software
configuration Combination type
Developing
experience
System
type
3 years
Well
equipment
windows
unix
10 years
incomplet e
equipment
CPT(A) CPT(C) CPT(P)
Figure 5. Information graph after relational connection
IV. REPRESENTATION OF KNOWLEDGE GRAPH
Knowledge is information that is structured and organized as a
result of cognitive processing and validation. Information is a
necessary medium or material for eliciting and constructing
knowledge. Knowledge is capable of being gained by learning.
For knowledge to be passed on entails encoding knowledge into
information and decoding again into knowledge. Knowledge
and information are not the same, but have as symbiotic
relationship [9]. Knowledge may be explicit for instance written
guidelines and implicit such as people’s experience and intuition.
The purpose of knowledge is to better our lives. In the context
of risk assessment, the purpose of knowledge is to reduce the
risk rate or to avoid risks as much as possible for the enterprise
and all its stakeholders. Knowledge representation and
reasoning formalism can also expression problems to be solved
concerning facts and general knowledge represented [11]. For
instance, one may ask with what kind of languages does Mike
speaking. Answering such questions requires descriptive
knowledge but also reasoning capabilities.
A. Abstraction on Knowledge Graph
Data and information are complex, from which we extract
valuable as a knowledge. Thereby we are capable of reducing
the available information capacity. When stakeholders obtain
the description of risk they are able to screen out the valuable
information and preserve those information as knowledge. As
for the example shown in Fig.5, decision maker will choose a
program with a lower rate of risk as Fig.7 shows. Page rank
algorithm works by counting the number and quality of links to
a page to determine a rough estimate of how important the
website is. The underlying assumption is that more important
websites are likely to receive more links from other websites
[15]. We adopt the idea of Page rank algorithm to filter useless

information and retain valuable information. Relevance is
measured as the probability that retrieved resource actually
contains those relations whose existence was assumed by the
user at the time of query definition. As Fig.7 shows, we give
each raw data contained in information a certain initial weight.
Ranks of information, denoted as R(Infor), can be calculated
according to (3):

   
where A indicates concepts and Ai indicates the raw data
elements of concept A. represents the weight of element Ai.
After calculating ranks of information, we can filter out
information that does not meet users’ query definition.
AW1
W2
A1
A2
BW3
W4
B1
B2
CW5
W6
C1
C2
A1
A2
B1
B2
C1
C2
W1*W3*W5
W1*W4*W5
W1*W4*W6
W1*W3*W6
W2*W3*W5
W2*W4*W5
W2*W3*W6
W2*W4*W6
Figure 6. Calculating ranks of information
Combination
type
Well
equipment U nix10 years 10%
30%
20%
10 years Well
equipment Windows
10 years Imcomplete
equipment Unix
Hardware and
software
configurat ion
Developing
experience
System
type
3 years
Well
equipment
windows
unix
10 years
imcomplete
equipment
Include
Include
include
include
include
include
Figure 7. Abstracted knowledge graph after filtering useless
information
B. Transformation on Knowledge Graph
With knowledge stakeholders can make more correct decision.
The context of knowledge graph can be created. Knowledge
graph can provide an open knowledge access interface and to a
certain extent it reflects the real world of inter-entity relations.
The graph structure of knowledge graph in Fig. 8 is not restricted
by form. Knowledge graph can express abundant natural
semantics and can supplement related information among terms.
The graph-based nature of knowledge graph makes possible a
linkage to other graphs thus resulting in an easy integrating of
multiple kinds of information and an enhancement in integrity
of information. By exploring the graph, new connections and
commonalities between items and users can be discovered and
exploited.
Combination
type
Well
equipment U nix10 years 10%
30%
20%
10 years Well
equipment Windows
10 years Imcomplete
equipment Unix
Hardware and
software
configurat ion
Developing
experience
System
type
3 years
Well
equipment
windows
unix
10 years
imcomple te
equipment
Include
Include
include
include
include
include
Bob
Unix
system
10 years
have skillIn
Figure 8. Supplementing semantic terms on knowledge graph
V. REPRESENTATION OF WISDOM GRAPH
Wisdom is an extrapolative process which includes
knowledge in an ethical and moral framework. Wisdom is the
process by which we can discern right from wrong and good
from bad. With wisdom we can judge from limited to infinity,
from known to unknown. Wisdom is the capacity to put into
action the most appropriate behavior, taking into account what
is known (knowledge) and what does the most good (ethical and
social considerations). Many informed people know what to do,
quite a few knowledgeable experts know how to do it, but only
a few wise persons know and can fully explicate why it should
be done. In line with these ideas the following metaphor applies
in Fig. 9, data: “know-who/when/where”, information: “know-
what”, knowledge: “know-how” and wisdom: “know-why”.
Wisdom as the ultimate unit of cognition is the result of
hierarchical processing of data, information, concept, and
knowledge. Knowledge is “knowing how” to do something,
wisdom is “knowing why, what and how” to do something.
Wisdom also extends to the application of knowledge in action.
A simplistic representation of the relationship between wisdom
and knowledge is captured in the following expression:
Wisdom=Knowledge + Ethics + Action. In Table 1 we take the
whole process of risk analysis of software engineering as an
example, and list the relations of data, information, knowledge
and wisdom in the process of risk analysis. At the initial stage of
the risk assessment, we collected some data about risk
assessment, which can be stored in arrays, stacks, lists, and so
on. Based on the data we collect, we can get descriptive
information about risk through conceptual mapping and
relational connection. And then according to the concept of
classification we store information in the relational database.
Identify applicable sponsor/s here. (sponsors)

Stakeholders can make a favorable decision after gaining this
risk description, and ultimately wisdom can help stakeholders
make future planning and forecasting.
Combination
type
Well
equipment U nix10 years 10%
30%
20%
10 years Well
equipment Win dows
10 years Imcomplete
equipment Unix
Hardware and
software
configurat ion
Developing
experience
System
type
3 years
Well
equipment
windows
unix
10 years
incomplete
equipment
Include
Include
include
include
include
include
Bob
Unix
system
10 years
have skillIn select
?
efficiency
?
predict
Figure 9. Using wisdom graph to predict unknown elements
TABLE I. EXAMPLE EXPLANATION IN DATA, KNOWLEDGE,
INFORMATION AND WISDOM ASPECTS
Aspects
Semantic load
Expression
Data
Input of risk
assessment
Array, list link,
stack, queue, tree,
graph
Information
Risk description
Relational
database
Knowledge
Understanding the
description and
make decision
Nodes and edges
with semantic
relations
Wisdom
(understand why)
the ability to use
results of the
analysis in the
right way
(frame or stylish
expression)
VI. PROGRESSIVE RELATIONSHIP OF DATA, INFORMATION,
KNOWLEDGE AND WISDOM
In Fig. 10, we show the progressive relationship between
data, information, knowledge, and wisdom through a pyramid
form. Data is one of the primary forms of information. It
basically includes recordings of transactions or events that will
be used for exchange between human or even with the machine.
Thus, unless the user understands the context in which the data
is collected, the data does not make sense. A word, a number or
a symbol can be used to describe a business result. It is the
context that gives data meaning, and this meaning makes data
informative. Information extends the concept of data in a broader
context. Therefore, information includes data but it also includes
all the information that a person associates as a member of a
social organization in a given physical environment. Information
like data is passed by symbol. These symbols have complex
structures and rules. Information has various forms, such as
writings, statements, statistics, charts or diagrams. When an
individual accepts and retains information as a true
understanding of reality and an effective explanation of reality,
the information becomes personal knowledge. On the contrary,
the organization or social knowledge exists when it is accepted
by the consensus of a group of people. Common knowledge does
not need to be shared by all members, and in fact it is accepted
by a group of insiders that can be considered a sufficient
condition. This is also the "public domain" knowledge of the real
[10]. Knowledge is a step further on the scale. It involves
understanding and ability to make use of the data and
information to answer questions, to solve problems, to make
decisions and so on. As the human mind uses this knowledge to
choose between alternatives, behavior becomes wise. Finally,
when values and commitment guide intelligent behavior,
behavior may be said to be based on wisdom. The level of
wisdom includes all the required components such as data,
information, and knowledge to make wise decisions.
Data
Information
Knowledge
Wisdom
Wisdom
graph
Knowledge
graph
Information
graph
Data graph
Why
How
What
Who/when/w here
Figure 10. Relationships among data, information, knowledge
and wisdom
VII. RELATED WORKS
Knowledge representation is a critical topic in AI, and currently
embedding as a key branch of knowledge representation takes
the numerical form of entities and relations to joint the statistical
models. However, most embedding methods merely concentrate
on the triple fitting and ignore the explicit semantic expression,
leading to an uninterpretable representation form [12, 13].
Traditional embedding methods do not only degrade the
performance, but also restrict many potential applications. In [7]
the authors proposed a semantic representation method for
knowledge graph, which imposes a two-level hierarchical
generative process that globally extracts many aspects and then
locally assigns a specific category in each aspect for every triple.
Because both the aspects and categories are semantics-relevant,
the collection of categories in each aspect is treated as the
semantic representation of this triple. In [14] the authors
proposed to represent knowledge in logical, philosophical, and
computational foundations.
VIII. CONCLUSION
Knowledge graph has been widely adopted these years, but
the expression of knowledge graph is usually limited to the form
of triples. We are increasingly aware of the semantic functions
of knowledge graph. In this paper, we elaborate the relationships

among data, information, knowledge, and wisdom, with the aim
of clarifying the expression of knowledge graph from the four
levels data, information, knowledge, and wisdom. And we
propose to specify the architecture of knowledge graph in the
four aspects of "data graph, information graph, knowledge graph
and wisdom graph". For users with complex information needs,
they can be allowed to express their needs by proposing natural
language questions. We can use the data graph to answer
questions asked by "Who / When / Where". Data are
meaningless in the absence of a given context. Information is a
combination of discrete data that gives an answer to the question
directed by "what". Knowledge is an effective combination of
abstracted and transformational information and capable to
answer questions guided by “what”. Wisdom is the ability to
criticize or act in a given situation. Wisdom can provide the
answer to "why" questions. Our work lays the foundation for a
survey from data to wisdom. In the next stage, we will deal with
data, information, knowledge and wisdom on the same
background and different backgrounds of the 5Ws problem and
explore more accurate expression of knowledge graph.
ACKNOWLEDGMENT
This paper is supported by National Natural Science
Foundation of China under Grant (No.61363007, No.61662021
and No. 61661019). * refers to the corresponding author.
REFERENCES
[1] Chaim Zins. Conceptual approaches for defining data, information, and
knowledge [J]. Journal of the American society for information science
and technology, 2007, 58(4):479-493.
[2] Bellinger G, Castro D, Mills A. Data, information, knowledge, and
wisdom [J]. 2004.
[3] Nadarajan G, Chenburger Y H. An Ontology-Based Conceptual Mapping
Framework for Translating FBPML to the Web Services Ontology[C]//
IEEE International Symposium on CLUSTER Computing and the Grid.
IEEE Xplore, 2006:8 pp.
[4] Rogers I. The Google Pagerank algorithm and how it works[J]. IEEE
Communications Letters, Vol: 2 Issue, 2002, 2(2):36--38.
[5] Lamberti F, Sanna A, Demartini C. A Relation-Based Page Rank
Algorithm for Semantic Web Search Engines[J]. IEEE Transactions on
Knowledge & Data Engineering, 2008, 21(1):123-136.
[6] Norta A, Ma L, Duan Y, et al. eContractual choreography-language
properties towards cross-organizational business collaboration [J].
Journal of Internet Services and Applications, 2015, 6(1): 8.
[7] Xiao H, Huang M, Zhu X. Knowledge Semantic Representation: A
Generative Model for Interpretable Knowledge Graph Embedding[J].
2016.
[8] Johannessen T V, Fuglseth A M. The effectiveness of data presentation
formats: An exploratory study[C]// Nokobit 2014, norsk konferanse for
organisasjoners bruk av IT. 2014.
[9] Lamberti D M, Wallace W A. Intelligent Interface Design: An Empirical
Assessment of Knowledge Presentation in Expert Systems[J]. Mis
Quarterly, 1990, 14(3):279-311.
[10] Syed A, Shah A. Data, Information, Knowledge, Wisdom: A Doubly
Linked Chain?[C]// International Conference on Information &
Knowledge Engineering, Ike 2006, Las Vegas, Nevada, Usa, June. DBLP,
2006:270-278.
[11] Chein M, Mugnier M L. Graph-based knowledge representation:
computational foundations of conceptual graphs[M]. Springer Science &
Business Media, 2008.
[12] Guu K, Miller J, Liang P. Traversing Knowledge Graphs in Vector Space
[J]. Computer Science, 2015.
[13] Minervini P, Fanizzi N, D'Amato C, et al. Scalable Learning of Entity and
Predicate Embeddings for Knowledge Graph Completion[C]// IEEE,
International Conference on Machine Learning and Applications. IEEE,
2015:162-167.
[14] Sowa J F. Knowledge representation: logical, philosophical, and
computational foundations[M]. Pacific Grove: Brooks/Cole, 2000.
[15] Lamberti F, Sanna A, Demartini C. A Relation-Based Page Rank
Algorithm for Semantic Web Search Engines[J]. IEEE Transactions on
Knowledge & Data Engineering, 2008, 21(1):123-136.