Research on the construction of three level customer service knowledge graph

Abstract

With the explosion of knowledge and information of the enterprise and the growing demand for intelligent knowledge management and application and improve business performance the knowledge expression and processing of the enterprise has become a hot topic. Aim at the problems of the electric marketing customer service knowledge map (customer service knowledge map) in building theory and method, electric marketing knowledge map of three levels of customer service was discussed, and realizing knowledge reasoning based on Neo4j, achieve good results in practical application.

Full text

IOP Conference Series: Materials Science and Engineering
PAPER • OPEN ACCESS
Research on the construction of three level customer service knowledge
graph
To cite this article: Shi Cheng et al 2017 IOP Conf. Ser.: Mater. Sci. Eng. 242 012077
View the article online for updates and enhancements.
This content was downloaded from IP address 168.151.117.55 on 28/09/2017 at 13:23

1
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd
1234567890
ICAMMT 2017 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 242 (2017) 012077 doi:10.1088/1757-899X/242/1/012077
Research on the construction of three level customer service
knowledge graph
Shi Cheng1,2, a), Jiajie Shen1, b) ,Quan Shi1,2, c) and Xianyi Cheng1,2, d)
1 School of Computer Science and technology, Nantong University, JiangSu Nantong
226019, China.
2 Nantong Research Institute for Advanced Communication Technologies, JiangSu
Nantong 226019,china.
E-mail: a) chenshi@ntu.edu.cn, b) 598864151@qq.com, c)sq@ntu.edu.cn, d)
Corresponding author: xycheng@ntu.edu.cn
Abstract. With the explosion of knowledge and information of the enterprise and the growing
demand for intelligent knowledge management and application and improve business
performance the knowledge expression and processing of the enterprise has become a hot topic.
Aim at the problems of the electric marketing customer service knowledge map (customer
service knowledge map) in building theory and method, electric marketing knowledge map of
three levels of customer service was discussed, and realizing knowledge reasoning based on
Neo4j, achieve good results in practical application.
1. Introduction
Power marketing refers to the power production, transportation and sales, and meets the power
customer’s economic and reasonable, safe and reliable use of electric products, and constantly
improves the floorboard of a series of economic activities of the economic benefit of power enterprise.
The power enterprise through appropriate and feasible marketing strategy, including quality service
strategy, price strategy, brand strategy, promotion strategy and other means to continuously improve
the electricity market share, in order to meet the need of customers, and to achieve the desired target
for power Enterprises [1]. In current society, the role and status of power marketing is increasingly
evident. Power supply electricity knowledge graph is the basis of power customers’ behavior modeling,
we did not find the relevant knowledge and power marketing knowledge graph from the open
literature [2].Knowledge modeling and application technology based on knowledge graph theory have
obtained and achieved good results the considerable development for many years at home and abroad
in many industries, but in the power industry of our country, the research and application remain to be
promoted. The knowledge graph can be used as the basis to explore the implicit relations among the
concept terms through logical reasoning, to make the tacit knowledge explicit and to realize the
sharing and reuse of domain knowledge.
Knowledge graph is an indispensable basic knowledge database for computer Natural Language
Processing, semantic search, computer reasoning and artificial intelligence. But there is no public
Chinese knowledge graph database, which has caused great difficulties to the relevant research.
Natural Language Processing, semantic search is quite developed in foreign countries, one is easier to
handle English than Chinese, and also English has many public knowledge graph database
(http://www.daml.org/onto- logiest/keyword.html [3]).

2
1234567890
ICAMMT 2017 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 242 (2017) 012077 doi:10.1088/1757-899X/242/1/012077
With the explosion increasing of enterprise knowledge and information, and the enterprise for
knowledge intelligent management and application is increasing, in order to facilitate the discovery
and application of knowledge and to improve business performance, the expression and processing of
enterprise knowledge have become the focus [4].
2. Related work
The concept of knowledge graph is proposed by Google in May 2012, and announced the basis for the
construction of the next generation of intelligent search engine. Although the concept of knowledge
graph is new, it is not a new field of study. In 2006 Berners-Lee proposed the idea of data link, thus
improved the relevant technical standards, such as URL (Uniform Resource Identified), RDF
(Resource Description Framework), WL (Web Ontology Language).Knowledge graph is based on the
relevant research results, and it is a revolution of the existing semantic network technology [5]
.The
knowledge graph that has been published and applied is shown in TABLE 1.
Table 1. The knowledge graph applied
Content
Baidu
knowledge
graph [6]
When Baidu search some key words of public figures, there will be the
relevant information of the character, the search results in the "Encyclopedia"
style are displayed. But now not only search popular characters, when users
search for names, subject or popular "facts", Baidu will give regular search
results on the left side of the search results, and search results on the right show
Baidu encyclopedia content with relevant keywords, and related search links.
Sogou cube [7]
Sogou said in its official micro-blog: in order to allow users to obtain
information more easily, Sogou search released a new knowledge base search
engine --“known cube”. This is the first search engine in the domestic search
engine industry.
Fudan GDM
Chinese
knowledge
graph [8]
Data mining analysis of micro-blog with knowledge graph to listen to
public opinion and improve people's livelihood research results, which are
published in the Liberation Daily, Xinmin Evening News and other newspapers
and reproduced by a number of network media.
Google
knowledge
graph [9]
In order to allow users faster and easier to find new information and
knowledge, Google search will release “knowledge graph” to search results for
the knowledge of the system, and through any key words can get a complete
knowledge system.
The relationship between the related concepts is given in figure.
Figure 1. The relationship between related concepts
As can be seen from Figure 1, ontology and knowledge graph is the basis of the construction of
knowledge system. Semantic web serves as a bridge between modeling and application. The
advantages and disadvantages of ontology knowledge graph have great influence on the quality of
graph database and knowledge base.

3
1234567890
ICAMMT 2017 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 242 (2017) 012077 doi:10.1088/1757-899X/242/1/012077
3. Three level architecture of power marketing customer service knowledge graph
3.1 Self-service architecture
Three levels knowledge graph used of power marketing customer service, providing customers with a
self-service customer service system (shown in Figure 2).Realize implementation of data extraction
from analysis to knowledge application.
Figure 2. Self-customer service based on customer service knowledge graph
Figure 7 shows the self-service customer service system architecture can be targeted for the user's
search information for context correlation, the relevant issues recommended the fuzzy intention of
intelligent guidance, flexible push key business, error correction and Pinyin analysis. The main
functions of the scene include business management, multi candidate output, sensitive word filtering,
intelligent input, voice recognition, context, customer care etc. Self-service customer service system is
low cost, 24h response throughout the day,the cost of a single service is 1/50 of an artificial hot line,
the effective diversion of 10%~20% artificial seat business, and the first time diversion of the business
can be maintained at more than 85%.
3.2 Application case analysis
This section mainly discusses the application of knowledge graph in power marketing customer
service. Of course, many application scenarios and ideas can be extended to other industries. The
application scene mentioned here is just the tip of the iceberg, and in many other applications, the
knowledge graph can still play its potential value.
3.2.1 Customer value analysis. Customer value analysis is a very important link in customer service.
The difficulty of analysis based on big data is how different sources of data (structured and
unstructured) to be integrated together. And which build the value analysis engine, so as to effectively
identify valuable customers.
A valuable customer is likely to be the beginning of a brand. A customer who has important value
is the key to a brand. The reputation of customers with important value directly affects and promotes
company's market influence and even market open, development. Having large customers can improve
the company's market value. An account for 20% of the total number of customers brings 80% of the
company's profit, which is the most important value of the customer to the company's most direct

4
1234567890
ICAMMT 2017 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 242 (2017) 012077 doi:10.1088/1757-899X/242/1/012077
impact on the value and benefits. In addition to the growth of corporate profits, the value of the
customer has a profound impact on the company.
And the important value of the customer will involve complex network, which also brings new
challenges to the value analysis. The knowledge graph, as a direct representation of the relationship,
can well solve these two problems. First, the knowledge graph provides a very convenient way to add
new data sources, as mentioned earlier. Secondly, the knowledge map itself is used to represent the
relationship, which can help us to more effectively analyze the specific potential value of complex
relationships.
Value analysis is a tool to identify the competitive advantage of enterprises. Value analysis pays
more attention to the value of customer service activities. Through the analysis of customer service,
the customer service activities are classified and analyzed, and the key link of customer value chain is
determined, so as to establish the source of competitive advantage.
The core of value analysis is customers, first need to put all the data source associated with a
particular customer to get through, and build the knowledge graph contains multiple data sources, so
as to integrate into a machine understandable structured knowledge. Here, we should not only
integrate the basic information (such as customer information when completing the application),can
also put the customer's demand information and behavior information, value information to integrate
the whole knowledge graph, thus reasoning.
3.2.2 New attribute inference.
Customer 1
Customer 2
Customer 3
Basic attributes
major
customers
Basic type
Outage sensitivity
Power quality
sensitive
Charge sensitive
demand
Low risk
Risk of
stealing
electricity
Electrical behavior
Winter
electricity
preference
demand
Excited
type
Technological
preferences
Energy-saving
service
Basic attributes
Big
customers
demand
expected
value
Security sensitive
Price sensitive
Electrical behavior
Seasonal
electricity
preference Summer electricity
type
Basic attributes
Customer
focus
Valley Type
Time
consume
preference
Electrical behavior
Figure 3. Power marketing knowledge graph reasoning
In Figure 3, the demand of customer 1 is “energy saving” and the demand of customer 2 is “billing
sensitive”, then we can the demand of customer 2 is also "energy saving”. Customer 3’s power
preference is “valley power type”, can also be inferred that the demand of customer 3 including
"energy saving”. According to the customer 2’s "steal electricity risk" low and customer 3 "security
sensitive”, can infer that the customer 3 is a potential customers.
The self-help service system based on knowledge graph can propose current knowledge graph for
user, which cannot answer content records and preliminary classification, to facilitate later artificial
intervention, the hot spot cared by new users added to the existing knowledge graph. With the
continuous improvement of user information in the customer service knowledge graph, according to

5
1234567890
ICAMMT 2017 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 242 (2017) 012077 doi:10.1088/1757-899X/242/1/012077
the operation characteristics and concerns of the user can intelligently remind the user whether needs
to query often concerned content, even users can regularly push attention information, gradually play a
service active perception and active service role in power marketing customer.
4. Conclusions
The knowledge graph is based on the new knowledge representation field of big data, is also the basis
of constructing knowledge base. There are many challenges in the exploration and practice of power
marketing customer service knowledge graph, this paper discusses three levels customers service
knowledge graph framework of power marketing.
The application of knowledge graph has a certain randomness, and the interpretation of knowledge
graph relies heavily on qualitative description and judgment, and cannot solve many practical
problems; The mapping of knowledge graph has certain technical barriers, and it is difficult for non-
experts to use it, It is a constraint factor that cannot have a significant impact on other disciplines and
social fields. Although some research makes a contrastive analysis of different data, methods and tools
of knowledge graph, but it did not form a specific application research standard, and the existing
research results are different or even contradictory, standardized data samples set for comparative
study is also a lack of.
In conclusion, the role of knowledge graph is not fully realized, and its conclusion is only auxiliary
verification, and the application needs to be further strengthened; At the same time, developing in
breadth, not only in the field of discipline (domain) structure visualization, but also in science and
technology management and scientific decision-making, enterprise innovation and competitive
intelligence play an important role. In addition, the application of knowledge graph is constrained by
the improvement of its theory and software tools; It is closely related to the development of knowledge
mining and artificial intelligence, and also needs to be combined with other semantic web technologies
to visualize applications; Much depends on a better unde-rstanding of the human brain. In a word, go
through the phenomenon of surface, it still has a long way to discover the trends and laws of subject
knowledge, and to visualize visually.
Acknowledgments
This paper is supported by the National Natural Science Foundation of China (No. 61340037),
Nantong Research Institute for Advanced Communication Technologies
References
[1] Junzhang Kang. Research on fine power marketing service management measures (Management
Science, China, 2017.09),pp.24-25.
[2] Guibin Wang. The Internet _ power marketing intelligent interactive service construction (China
power enterprise management, China, 2017.01),pp.48-49.
[3] Xiao Tian, Yongchao Liu,Jing Wang, et al. Application value of customer service knowledge graph
construction of Power Grid Corp(Shandong electric power technolo,China,2015,42(12)),pp.65-69.
[4] Changle Xu. Analysis of power network marketing mode under smart grid (communication world,
China, 2017, 03),pp.219-220.
[5] Jiao Liu, Yang Li,Hong Duan, et al. Overview of knowledge graph technology(computer research
and development,China,2016,53(3)),PP.582-600.
[6] Baidu next generation search engine prototype exposure application knowledge mapping
technology (Computer programming skills and maintenance, China, 2013,19),pp.4-4.
[7] Sogou cube.http://baike.sogou.com/v66616234.htm.In Chinese.
[8] Wei Li,Yanghua Xiao,Wei Wang. Character entity recognition based on Chinese knowledge graph
(computer engineering, China, 2017,43(3)),pp.226-234.
[9] Ochs C, Tian T, Geller J,et al. Google knws who is famous taday- building an ontology trom search
engine knowledge and DBpedia[C]//Proc of the 5th IEEE Int Conf on Semantic
Computing.Piscataway, NJ:IEEE, 2011:320－327.

6
1234567890
ICAMMT 2017 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 242 (2017) 012077 doi:10.1088/1757-899X/242/1/012077
[10] Yisong Ma,Zhigang Wu.Modeling and analysis of large power data based on Neo4j(new
technology of electrical and electrical energy,China,2016,35(2)),pp.24-31.