See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/334232412

Development of an Expert System-Oriented Service Support Help Desk

Management System

Chapter · January 2020

DOI: 10.1007/978-981-13-7564-4_57

CITATIONS READS

2 2,631

5 authors, including:

Abrar Hasin Kamal Mohammad Obaidullah Tusher

Universität Potsdam University of Koblenz and Landau
5 PUBLICATIONS 47 CITATIONS 2 PUBLICATIONS 6 CITATIONS

SEE PROFILE SEE PROFILE

Nusrat Jahan Farin Nafees Mansoor

Stamford University Bangladesh University of Liberal Arts Bangladesh
13 PUBLICATIONS 90 CITATIONS 101 PUBLICATIONS 919 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Mohammad Obaidullah Tusher on 11 June 2021.

The user has requested enhancement of the downloaded file.

 Development of an Expert System Oriented Service
Support Help Desk Management System

Abrar Hasin Kamal1, Mohammad Obaidullah Tusher1, Shadman Fahim Ahmad1,

Nusrat Jahan Farin2 and Nafees Mansoor 1
1
University of Liberal Arts Bangladesh (ULAB), Dhaka, Bangladesh
2
Stamford University Bangladesh, Dhaka, Bangladesh
[email protected], [email protected], [email protected],
[email protected] and [email protected]

Abstract: With the recent advancements of artificial intelligence, automation has

been making its way into various systems that previously could only be operated
by humans. Hence, these systems are getting more autonomous and taking the
control off from the human operators gradually. On the other hand, service cen-
ters are found to be subunits of almost all companies. However, a process at such
service centers becomes slower and inefficient once the center is completely op-
erated by humans. On the contrary, considering the functionalities of the center,
complete automation is not either effective in certain centers. This triggers the
necessity of a semi-automated center where a framework for such service center
is proposed in this paper. While analyzing, it has been observed that the perfor-
mance of the existing system is heavily hampered due to the absence of dy-
namicity. Therefore, the proposed framework attempts to make the center dy-
namic by introducing certain automated behaviors. The proposed system dis-
cussed in this paper comprises of all features in that are present in the current
system but in an automated way. Furthermore, the proposed system extends the
functionalities of a service center that could not be achieved by humans only. The
proposed system also incorporates AI to a part of it making it semi-AI driven. A
further description is also given in this paper of what needs to be done in order
to accomplish a completely AI-driven system.

Keywords: Expert System Automation, Service Support Center, Artificial In-

telligence, Semi-AI Model, Call Support Framework

1 INTRODUCTION

Authorized physical establishments that provide services, free of charge (if within
the period of warranty), such as repairs and replacements are known as service centers.
Service centers are established as sub units of companies to rake in monetary profit and
provide support to stakeholders even after they have sold their product. This makes
service centers an ongoing process which obtains the costs for its operations by charg-
ing users in exchange of a service. A sub unit of service centers are repair centers that
consist of skilled technicians on the domain of the products an organization sells.
2

The process of dynamically selecting data, converting information, deciding and ex-
ecuting accordingly is known as Automation [1]. One of the major trends of the 20th
century has been the surge of enabling electro-mechanical systems to take more control
of the task they are designed to do. This has been done from the belief that automated
systems can provide increased reliability, better performance while being cost-effective
in context of various functions [2]. Automated systems have been around for a consid-
erable amount of time now. They serve the purpose of making our everyday tasks easier
and efficient. Automated systems make decisions based on inputs to carry out further
work that it is intended to do, just like humans. Automation doesn’t only limit a system
from delivering the expected output but also extends the functionalities of a system.
This extensively often exceeds the capabilities of humans, for which the shift towards
automation has been so visible in the recent past. However, the term is yet to be com-
pletely justified as a human actor remains an integral part of an automated system. In
most cases, this is due to the responsibility of monitoring failures of the system [2].
Automation of service centers opens a horizon of opportunities for companies to
improve the quality of service they provide to their customers while scaling down on
their costs that is spent on the resources, both human and logistical, of a service center.
Automated service centers would also benefit the companies in terms of the revenue
they obtain from it as an automated system would be more dynamic in terms of pro-
cessing ongoing tasks and their status. This would allow a greater number of customers
to be served, hence, increasing profit.
AI can be used to manage users and the system that is used to get them the help that
they seek regarding their problems. It can also predict with utmost accuracy of the re-
quirements of the user’s problems. Intelligent auto responders may also be used to au-
tomatically provide users with a response email containing a range of the most probable
solutions. When a solution works through an automatic embedded button presses re-
cording the solution that worked and updates the status of the issue. AI can also be
made use of in the generation or looking up of new or existing articles respectively [8].
Service centers mostly operate through manual methods which consist of several
methods that lack dynamicity. Customers are required to report their problems to a staff
member of the center. A manual check is usually done to check if the hardware is in
within the period of its warranty. The next step is to assess the hardware to identify the
issue and then forward it to the appropriate technician for servicing. Furthermore, the
customers are not aware of the status of their product that is being repaired. This makes
the current workflow of service centers inefficient. In this paper, we propose an expert
system-oriented service support help desk management system which is partially driven
by Artificial Intelligence and automates the current work flow. We say partial due to
the fact that human interaction is still required for assigning the keywords.
This paper is organized as follows. In section 2, the background literature is pre-
sented where strengths and limitations of the existing service centers are discussed. In
Section 3, the proposed model of the expert system is explored and the implementation
framework is discussed in Section 4. The paper ends with section 5 where the future
scopes and conclusion are highlighted.
 3

2 BACKGROUND STUDY

Companies have been using various Enterprise Resource Planning (ERP) tools for
maintaining their service centers. Tools such as Local ERP Desktop Base, Soft Neck
ERP and Business Root ERP Software are used for the operations of the service centers.
In all the current systems that are being used in service centers a common transportation
problem persists. Customers are required to carry their hardware to the service centers,
and in cases of heavy hardware such as monitors and CPUs, this is often problematic
[3].
The Local ERP Desktop Base software only provides warranty support. The system
is capable of checking if a product is covered by warranty or not [3]. The system also
uses the products barcode to generate an invoice of the product. The staff concerned
with all the aforementioned steps then listens to the problem being faced by the cus-
tomer and forwards it towards the technician who specializes on the issue. Customers
are often unfamiliar with technical terminologies and may fail to properly identify the
hardware issue which leads up to an extension of the timeframe that is required. The
local ERP Desktop base is exclusively accessible to only one of the actors in the whole
system.
Soft Neck ERP is another tool that is widely used in organizations to manage the
operations of their service centers. Companies using Soft Neck ERP require their cus-
tomers to first report to an administration focus. In soft neck ERP, the service for ad-
ministrative moderation of the product is done free of charge, if the hardware is within
its warranty [4]. A possibility of the administration for a product under service being
expelled exists in soft neck ERP. This is done for various reasons, but mainly occurs
due to inconclusive timeframes [5].
The Business Root ERP software is also an enterprise resource planning tool that is
widely used by companies worldwide. The Business Root ERP tool is an online base
where the record of a products warranty is stored when it is sold [6]. This system is also
exclusively open to only one of the actors in the whole system. Companies using this
system require their customers to carry their products to the service centers which prove
to be a tedious task. After-sales service have been extensively researched on in the past,
but a complete framework is yet to be developed. Studies on after-sales have put for-
ward the importance of a strategic after-sales methodology [6][7].

3 PROPOSED MODEL

The design and development of the automated service center has been done keeping
the gaps in the existing system in mind. The key features of this system set it apart from
the rest as the system developed it is a semi AI system. The flowchart of the system is
illustrated in Fig. 1.
4

Fig. 1: System flowchart of automated service center.

3.1 Dynamic

This paper proposes the design and development of an automated service center that
implements a framework fulfilling key after-sales gaps that are present in the existing
operation methods of service centers. Automation of the process adds dynamicity to the
whole process [12][13][14]. It also extends functionalities of the system which couldn’t
have been achieved by human. The involvement of all three actors, customers, service
providers and the system itself, adds more dynamicity to the system.

3.2 Customer’s involvement

The system allows users to have their personal accounts which they get upon com-
pletion of registration. Customers gain access to their respective dashboards from where
they can submit a request for service. A gap in the current existing framework is the
unavailability of information regarding a product’s status. The dashboard also notifies
users about the status of their product that has been submitted for service.

Algorithm 1: Keyword Scraping Algorithm for the Proposed System.

 5

1: Import libraries (sys, bs4 as bs, requests, newspaper)

2: Set search_text = Keywords // User Problem Centric
3: Set url = “https://www.google.com/search?q=” + search_text
4: Set page = Retrieves the contents of the url using Get Request
5: Set soup = Retrieves data from html page
6: Set links = Retrieves the links from extracted data
7: Set FINAL_LINKS = []
8: Loop
9: Set short_link // In order to find “a” tag
10: if (short_link.find(“youtube”) = = -1)
11: Set start_string = finds “http” in short_link
12: Set end_string = finds and sets “&” in short_link
13: FINAL_LINKS.append(short_link[start_string:end_string])
14: Else Step 9
15: Prints FINAL_LINKS
16: Set FINAL_KEYWORDS = []
17: Loop
18: Set article_url = each website link in FINAL_LINKS
19: Set a = Article(article_url, keep_article_html=True)
20: a.download()
21: a.parse()
22: Set html_article = a.article_html
23: a.nlp() uses natural language processing for keyword tokenization
24: Set keys = The keywords found in the html document
25: Set FINAL_KEYWORDS = Adds the keys to the FINAL_KEYWORDS
26: Prints FINAL_KEYWORDS

3.3 Keyword Search

A unique keyword search system is proposed in the system that eliminates the as-
sessment procedure which can be very time consuming [11]. In Section 2, it has been
mentioned that in all current existing systems, the staff receiving the product for service
has to listen to the user for identifying the problem. This is an error-prone procedure as
customers may not be familiar with specific technical terms that define their hardware
issue. This may often lead to a longer amount of time being spent behind assessing the
problem. While submitting a service request, customers enter a keyword that represents
the issue they are facing. A background search executes that scrapes the web to find a
set of words that are related. This keyword search mechanism uses its result to match
the appropriate technician who is specialized in the word that has been searched for.
The instructions followed by the system are described in the algorithm below.
In machine learning the procedure of predicting the class of given data points is re-
ferred to as classification. Lazy learners and eager learners are the two types of learners
found in classification. A lazy learner is one that waits for data to appear and stores the
training data. Once the data appears the most related data found in the stored training
data is used for classification. One such lazy learner algorithm used for the expert sys-
tem is the k-nearest neighbor (kNN) algorithm (Fig. 2). All instances which are propor-
tionate to the training data points are stored in a n-dimensional space by the algorithm.
When the system receives a distinct/discrete data is found the nearest k instances (near-
est neighbors) are considered and analyzed after which it provides the most common
6

class as the prediction. The mean of the kNN is provided in the case of real-valued data
[9].

𝟏
𝒘 ≡ 𝒅(𝒙 𝟐
Equation (1)
𝒍 , 𝒙𝒎 )

To compare the distance between each point from the input the aforementioned
equation is put to use, where d is the distance between the two points xl and xm. The
future prospect of the expert system holds an AI driven system which will make use of
a deep learning model such as a Convolutional Neural Network (CNN) model. The
model will ease the process of user input, which will now be images instead of text.
This is due to the reason that CNN has been developed mainly for the purpose of image
classification.

Legend
Word classification
Word classification

Search query
?

Fig. 2: Illustration of the methodology of kNN

3.4 Ease of Transportation

The background literature highlights the inconvenience customer’s having to face
while carrying their products to and from the service center. The proposed system elim-
inates this problem by integrating a delivery module in the system. After the service
request is accepted and the product has been submitted for service, customers can mon-
itor the status of their product. Once their product is ready for delivery, the companies
can forward the product to the delivery team and notify the customers through the dash-
board. This exempts customers from taking the hassle of coming over to the service
centers to receive their product and carrying it back home.

4 FRAMEWORK IMPLEMENTATION

4.1 Architectural Design

The architectural design of the proposed system uses a three-tier approach to inte-
grate all the actors and their functionalities on one platform. The following illustration
represents the three-tier architecture for the automated service center.
 7

Fig. 3: The 3-Tier Architectural Design of the system

Fig. 3 is an illustration of the 3-tiered architecture of the proposed automated system.

The first tier is the Client Tier followed by the Middle Tier. The third tier comprises of
the database.
The Client Tier: The first tier in the architecture is the client tier which allows
users to interact with the platform. The web browser in this case acts as the inter-
face for the client tier. The web browser is responsible for processing and display-
ing the HTML resources, problems and communication protocol requests. The
CodeIgniter framework of PHP has been used for development of these pages for
faster operations. The client tier comprises of the functions a client/technician can
perform, such as signing up, logging in, submitting a service request and so on.
Fig 4 shows the service request form a customer is required to fill in to submit a
service request.
The Client tier is utilized by all three actors: Engineers, Customers and Admins,
of the system. Both admins and engineers have a similar dashboard with certain
privileges depending on their roles.

Fig. 4: Service Request Form

The Middle Tier: The middle tier of the proposed automated system comprises of
most of the application’s core logic. The client tier, as mentioned earlier presents data
and collects data from the user and the database tier is responsible for storage and re-
trieval of the data. The rest of the functionalities lie in the middle tier. The middle tier
serves as a merger between the other tiers. The middle tier is responsible for processing
8

a user’s input and rendering an output accordingly. Inputs coming from the user are
formed into queries which the database interpret to read or write data. The proposed
model in this paper is said to be semi-AI driven. The AI part of the application is also
present in the middle tier. A unique functionality that has been discussed in section III,
which assigns technicians to service requests depending on keywords requires AI to
cluster like-words together. The web server, web scripting language and the scripting
language engine are the core components of the middle tier.
The Database Tier: The third and last tier of the proposed system is the database
tier. This tier is responsible for the storage and retrieval of data related to the system.
This tier consists of five database tables.

4.2 Roles of Actors

The proposed system comprises of three types of users. Regardless of the type of
user, all actors of the system must be registered on the platform to use it. All users are
redirected to their respective dashboards upon successfully logging in. If a customer
logs in, they are prompted with the options to submit a request for service. Customers
are also able to make payments through their accounts before the service center starts
repairing the product. The admin’s dashboard allows extended functionalities such as
viewing all requests from all users, adding new engineers and updating their infor-
mation and adding sold products to the sold product list. On an engineer’s dashboard,
engineers can set the product status that are under their service. They can also view the
service requests waiting to be accepted.
Customer: The first role of a User in service center is to open an account and to
login into the account via email. Next, user sends request for a product service or war-
ranty and tracks history. After the system performs a warranty check on the product it
lets the user know. The customer is redirected to either the keyword search page or the
payment page, depending on the product’s warranty status. Customers are required to
complete payment if their product does not have any warranty. All customers are re-
quired to identify their issue using a keyword. Given the fact, that all customers may
not be able to identify their issue using the appropriate keyword, a background script
scrapes the internet and fetches a set of five to seven words that are related. The scrap-
ing [10] is done from the top links suggested by Google for the word that has been
searched by the customer. Considering the user’s role and activities there are a number
of features in the proposed automated service center. Some of the requirements are (a)
Creating a new account to access the system (b) Logging into the system (c) Sending
service request (d) Payment option (e) Logout
Admin: The admin plays a very important role in the proposed system. The admin
core task is to manage the database. Admins are also responsible for keeping track and
maintaining records. Admins are also given the authority to view all service requests,
add products to sold list and insert or edit engineer’s information. The customer re-
quests for registration on the platform are also subjected to be approved or disapproved
by an admin. Admins do not require the process of registration, thus can directly log in
to the system. Considering the roles and activities of an admin, some of the require-
ments of an admin’s role are (a) Logging in (b) View all current and previous service
 9

requests (c) Insert Engineer Info (d) User Registration Request (e) Insert Sold Product
information (f) Log out.
Engineer: The first role of an engineer in the automated service center system is to
login to the system using their registered email address. Upon succession, engineers
can update the current status of the products that are under their service. They can also
check their service history to refer back to any issue they have already solved. The
current requests that are yet to be accepted are also displayed on the engineer’s dash-
board. Considering the engineer’s roles and activities in the proposed system, some of
the requirements of an engineer are (a) Log in (b) View current requests (c) Update
status (d) Help.
Physical Design: Upon completion of the system’s analysis and understanding the
role of each functionality of each actor, the logical design of the system was created.
Using the help of Data Flow Diagrams (DFD) each process was modeled. An Entity-
Relationship diagram (ERD) was constructed to better understand the flow of data
through the system. Sequence diagrams were also used to illustrate the roles and func-
tions of each actor. The physical design of the system solely focused on the workflow
of the system and its resulting output. Fig. 5 and Fig. 6 depicts the sequence diagrams
signing up and logging in.

4.3 Application
Regardless of the type of user, all actors of the system must go through the process
of logging in. Depending on the type of user that is logging in, the dashboard will dis-
play with the corresponding functionalities as depicted in Figure 4. Sequence diagrams
of signing up and signing in have been shown in Fig. 5 and Fig 6.
The process of signing up is fairly simple. Customers are prompted with a form after
they have clicked on the sign-up button. They are required to fill out mandatory infor-
mation in the form and click on submit which sends the details of the customer to the
database. Once they have completed this process, they are ready to use the platform.
Signing in the system is a process that is common for all users of the system. Users are
required to enter their credentials in the log in form.
A key feature of the proposed system is the semi AI driven keyword searching sys-
tem. Before submitting a service request, customers are required to enter a keyword
identifying their issue. A python script scrapes the web for related words that can iden-
tify the problem better. After the set of words are retrieved, the AI clusters like words
together to identify the problem and assigns the technician with the specialization that
best fits the problem. Fig. 7 shows the sequence diagram for the keyword search feature.
The script searches the keyword entered by the user in Google and returns a list of
words that comes up in the first few links. A portion of Fig. 8 represents the output
shown after a word was searched.
10

Fig. 5: Sequence Diagram for customer sign up

Fig. 6: Sequence Diagram for user sign in

 11

Fig. 7: Keyword search sequence diagram

A real world implementation of the system may be described as a user looking for a
solution for the problem of “Slow RAM”. After signing in to the system the user is
prompted to enter the keywords that defines their problems, in this case the keywords
entered is “Slow RAM”. The output received as a result of the keyword entered by the
user can be seen in the Fig 8. Once the array of words has been loaded, it is fed into the
kNN model. The semi-AI aspect of the system is found at this portion of the system.
Taking the array of words, the model calculates the distance of each word against the
technical terms it already knows using Equation (1). The word with the highest score is
then searched for in the database. Once the word is received, the technician whose
skill/specialization matches with that of the matched word is then assigned to provide
the user with a solution of “Slow RAM” and what they need to do in order to get that
solution working. As mentioned earlier, a number of checks follow up to ensure that
the technician selected is currently available to attend to the user’s request. To maxim-
ize the pool of keywords, the system stores matched keywords. This expands the
knowledge base of the system and makes the keyword search feature more reliable.
12

Fig. 8: Results after keyword searching

5 CONCLUSION AND FUTURE WORKS

An automated, semi AI based, service center system is proposed in this paper. The
proposed system adds dynamicity with the involvement of all types of users as well as
real time information. This proposed system also adds more flexibility to the whole
process through elimination of the need for customers to physically be present in order
to submit or receive their products for service. Moreover, the proposed methodology is
subjected to be beneficial for both customers and the companies, as service centers can
attend to more customers and their requests, hence making profit. Future works of the
proposed model mainly focuses on the implementation of the AI that assigns service
requests to engineers.

REFERENCES
1. Lee, J.D. and See, K.A., 2004. Trust in automation: Designing for appropriate reliance. Hu-
man factors, 46(1), pp.50-80.
2. Endsley, M.R., 2018. Automation and situation awareness. In Automation and human per-
formance (pp. 183-202). Routledge.
3. Panizzolo, R., 2008. A methodology to measure the value of services provided to customers
in manufacturing firms. Measuring Business Excellence, 12(3), pp.3-15.
4. Hewitt, P., 2002. The government’s manufacturing strategy. Secretary of State for Trade
and Industry, Westminster, 4.
5. Oliva, R. and Kallenberg, R., 2003. Managing the transition from products to services. In-
ternational journal of service industry management, 14(2), pp.160-172.
 13

6. Karlsson, U., 2007. Service Based Manufacturing Strategies Implication for Product Devel-
opment, Production and Service Operations in Global Companies.
7. Neely, A., 2008. Exploring the financial consequences of the servitization of manufacturing.
Operations management research, 1(2), pp.103-118.
8. Riley, K., McVicker, W., Nunn, S., Anand, S. and Brett, J., Accenture Services Ltd and
Accenture LLP, 2002. Method for implementing service desk capability. U.S. Patent Appli-
cation 10/029,769.
9. Sebastiani, F., 2002. Machine learning in automated text categorization. ACM computing
surveys (CSUR), 34(1), pp.1-47.
10. Lawson, R., 2015. Web scraping with Python. Packt Publishing Ltd.
11. Arasu, A., Cho, J., Garcia-Molina, H., Paepcke, A. and Raghavan, S., 2001. Searching the
web. ACM Transactions on Internet Technology (TOIT), 1(1), pp.2-43.
12. Parasuraman, R., Sheridan, T.B. and Wickens, C.D., 2000. A model for types and levels of
human interaction with automation. IEEE Transactions on systems, man, and cybernetics-
Part A: Systems and Humans, 30(3), pp.286-297.
13. Farin, N.J., Rimon, M.N.A.A., Momen, S., Uddin, M.S. and Mansoor, N., 2016, December.
A framework for dynamic vehicle pooling and ride-sharing system. In Computational Intel-
ligence (IWCI), International Workshop on (pp. 204-208). IEEE.
14. Farin, N.J., Rahman, A., Mansoor, N. and Hossain, S., 2016. Wotcoms: A novel cross-lay-
ered web-of-things based framework for course management system. In ICAICT.

View publication stats