Proceedings of the Sixth International Conference on Trends in Electronics and Informatics (ICOEI 2022)

IEEE Xplore Part Number: CFP22J32-ART; ISBN: 978-1-6654-8328-5

Automated Resume Screener using Natural

Language Processing(NLP)
Tumula M ani Harsha Gangaraju Sai M oukthika Dudipalli Siva Sai
Deparment of CSE Deparment of CSE Deparment of CSE
SRM University-AP SRM University-AP SRM University-AP
Amaravati, Andhrapradesh Amaravati, Andhrapradesh Amaravati, Andhrapradesh
India India India
[email protected] [email protected] [email protected]
2022 6th International Conference on Trends in Electronics and Informatics (ICOEI) | 978-1-6654-8328-5/22/$31.00 ©2022 IEEE | DOI: 10.1109/ICOEI53556.2022.9777194

M annuru Naga Rajeswari Pravallika Satish Anamalamudi M uraliKrishna Enduri

Deparment of CSE Deparment of CSE Department of CSE
SRM University-AP SRM University-AP SRM University-AP
Amaravati, Andhrapradesh Amaravati, Andhrapradesh Amaravati, Andhrapradesh
India India India
[email protected] [email protected] [email protected]

Abstract—Resume S creening is the process of evaluating Screener” starts working by scanning all the resumes given
the resume of the job seekers based on a specific requirement. to it and provides an output regarding the skill sets of the
It is used to identify the candidate eligibility for a job by candidates by which the eligible candidates for the
matching all the requirements needed for the offered role with requirement can be chosen easily. “Resume Screener” is
their resume information such as education qualification, skill basically an automated resume recommendation system,
sets, technical stuff etc. Resume S creening is a crucial stage in which is based on the Machine Learning techniques[7,8].
candidate’s selection for a job role, it is the stage where the This model takes the input in Word/PDF format and reflects
decision making is done whether to move the candidate to the
the output in the PDF, Image and CSV formats. The output
next level of hiring process or not. Traditionally, this process is
will be displayed in the form of graph which includes points
performed manually, but companies often receive thousands of
resumes for job applications. In order to reduce the human
for each skill which are required for the job profile[9-12].
involvement and errors, many new ways were introduced in
According to the output the eligible candidates for the
this process. This paper discusses about one such process required job can be selected easily. “Resume Screener” is
which is very efficient in performing Resume screening. It very efficient and less time consuming. It is also practically
includes Natural Language Processing (NLP), an automated possible in the current situations[13-16].
Machine Learning Algorithm for screening the resumes. This
paper explains the end to end working of a python application Resume Screening(Resume Shortlisting)
which efficiently screens the resumes of the candidates based
on the organization’s requirement. Technical Written Test/Tests

Keywords— NLP, Resume Screening, Hiring Process, Skill Technical, Behavioural, Manageral
Set.
Interviews
I. INT RODUCT ION HR Interviews
Effective screening of “Resume” plays a pivotal role in
current scenarios due to increased volume of the received Offering the Job
applications[1,2]. This screening process requires more
domain knowledge to understand the requirement and
eligib ility for the job role. A large number of job roles are Figure.1. Basic Job Hiring Process
existing currently for which different technologies are
required. In these situations, huge number of candidates are Select resumes that have the required
applying for the jobs in which the eligible candidates should credentials
be selected for which the basic process of Resume Screening
should be done[3-6]. Resume Screening is mainly used for Select resumes that have the desired skills
the separation of right candidate for the required profile. So,
the screening process should be efficient, for which “Resume
Screener” can be used. This “Resume Screener” is designed Select resumes that are customized for the job
and built by using the Machine Learning approach
commonly known as “Natural Language Processing”.
Figure.1 exp lains about the job hiring process which Check the applicant's information
basically works with five different steps. Subsequently,
Figure.2 explains about the the steps involved in sceening the
resumes during the process of job hiring. “Resume

978-1-6654-8328-5/22/$31.00 ©2022 IEEE 1772

Authorized licensed use limited to: ULAKBIM UASL - CUKUROVA UNIVERSITESI. Downloaded on November 09,2022 at 11:21:25 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Sixth International Conference on Trends in Electronics and Informatics (ICOEI 2022)
IEEE Xplore Part Number: CFP22J32-ART; ISBN: 978-1-6654-8328-5

Figure.2. Basic Process of the Resume Screening The basic working process
The resume should be in the PDF format then the resume
II. CHALLENGES IN RESUME SCREENING can be opened, read and extract the text. Later the extra text
a.Time Consuming-Higher level of difficulty if the no. of will be removed and the keywo rds like requirements will be
resumes is high termed area wise. The process gets continued by calculating
b.Quality of Hiring-The quality of hiring gets lower when the scores per each area and gets sorted for final scores
there is bulk of resumes. creation. Finally based on the scores a pie chart will be
c.Hiring Biases-The recruiters can be biased towards displayed as output by which the recruiters can select the
some candidates. required and eligible candidates for the offered job role[20].
d.Recruiter’s Experience-If the recru iter is not Problems
experienced about all the skills required for the job then the Time Complexity
screening process may not be efficient. Scanning each resume at once and calculating the skills
e.Recruiters Search-Recruiters may stop the search if the takes much time.
suitable candidate is found and do not go through the entire Not user friendly
stack. Only the programmer can use the application efficiently.
f.Unnecessary allocation of recruiters -Separate recruiters Tightly coupled blocks
should be placed for the resume screening process which Every code block is dependent on each other to bring any
results in waste of time and resources[17]. change the whole code should be disturbed which may
break the flow [21].

III. EXIST ING SOLUT IONS AND PROBLEMS

C. Resume Screening using Machi ne Learning
Classifiers
a. Manual Screening
Manual Screening of resumes is nothing but some of the This was built using Machine Learning algorith ms and
employees of the company which is going to provide the job classifiers to screen the resumes[22].
will involve in the resume screening process i.e. manually
they check each and every application of the candidates The basic working process
resumes one by one till they find their requirement. It can be The resumes should be in the CSV format and then the
done based on the skills they want or work experience of the screening process starts with the junk words
candidates or some other categories which are suitable for (unwanted/repeated words) removal fro m the resumes. Later
the job profile. the remaining wo rds get screened and points for the skills
Problems get awarded. And the skills points will be sorted
Time Consuming accordingly[23-24]. Finally based on the skill points there
All the resumes should be referred manually wh ich takes so will be a graph displayed as the result by which the eligib le
much of time. candidates for the job role can be selected.
Unnecessary allocation of the resources Problems
Recru iters can involve in any other developments instead of Format Issues
spending so much of time into this resume screening. Only CSV format resumes will be considered which is not
Inefficient possible in all situations as the main fo rmats for resumes
Recru iters don’t go through all the resumes once they find will be Word or PDF.
their requirement. Very Complex approach
Biased Idea of dealing with these sorts of approaches will be
The recru iters may be biased towards some o f the complex and cannot be understood by everyone[22].
candidates as per their convenience [18]. Tightly coupled blocks
Every code block is dependent on each other, to bring any
b. Resume Screening using Artificial Intelligence change the whole code should be disturbed which may
break the flow[23].
Artificial intelligence, along with text min ing and Loss of data
processing algorithms is used to develop these types of By using some approaches like “genism” (library used for
Resume Screening So ftware. These programs usually look filtering the given data) there might be loss of some
for specific key words to sort resumes and rank them to important data which might result in the efficiency of the
determine the job applications that should be further engine.
reviewed by recruiters[19].
d.Resume Screening using Deep Learning

978-1-6654-8328-5/22/$31.00 ©2022 IEEE 1773

Authorized licensed use limited to: ULAKBIM UASL - CUKUROVA UNIVERSITESI. Downloaded on November 09,2022 at 11:21:25 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Sixth International Conference on Trends in Electronics and Informatics (ICOEI 2022)
IEEE Xplore Part Number: CFP22J32-ART; ISBN: 978-1-6654-8328-5

Built using the Deep Learning approach and it works by

learning from its previous data. b.Functional Overview
The basic working process The whole flow is divided into four parts.
Category: Type of Job Resume fits for. •The User interface consists of two buttons along with so me
Resume: Resume of candidates other info, in which one of the buttons act as format
By using these category and resumes provided, the transformer which transforms files (in any other formats) to
screening begins. It intakes each resume and screens it and pdfs. The second button is the key to start the real process
generates output to that particular resume by using natural which handles the request to screen the pdfs. As soon as this
language processing by resulting in a pie chart. A different button was hit, the flow starts and the application screens
graph will be provided per each resume. It will be getting the resumes.
efficient per each run. •The folder part of the application consists of four folders in
Problems which one is the input folder where resumes are to be
No Bulk resume screening loaded, the other is the destination folder in wh ich the
Only screens one resume per each run and generates output resumes transfer after the screening is done, the other one is
for the single resume. the unscanned folder in wh ich the resumes which are no t
Time Taking scanned properly due to the fo rmat issues are stored and the
Higher the resume number, higher the time of screening. last one is the output folder where outputs are stored. This
folder approach makes it easy for the user to distinguish the
IV. PROPOSED SOLUT ION screened and unscreened resumes.
The proposed solution is designed using Natural Language •The third part of the application is code part which handles
Processing (a Machine Learning approach) and the the requests from user and performs the process. The
Application is named as “Resume Screener”. Th is model application server should be up and running in order to
takes the input in Word/PDF format and reflects the output handle the requests from the user.
in the form of PDF, Image and CSV formats. The goal of •The fourth part is output part, it is also a folder in which the
the project is to screen the resumes without any manual output of the application is stored in three formats, PNG,
intervention and provide the efficient result. To achieve this PDF and CSV. Based on the requirement of the user any of
goal, a python project is designed as a web application the format can be used. Upon each run, these files will be
which can be triggered by APIs. overridden and replaced by the new outputs.
•The fifth part is the Dataset part. It is a CSV file which
a. Architectural Overview: consists of various skills under various roles. This is used by
The architecture of this project consists of five blocks as the application to screen and calculate the frequencies of the
shown in Figure.3. resumes. The output graph consists of different frequency
1.User Interface: Fro m where the requests are made to the points for different skills wh ich will be calculated by
application. comparing the skills of the candidates mentioned in the
2.Application: The part which handles the requests from resume and required skills for the job profile mentioned in
user interface and performs business logic. the data set. This data set can be updated with different and
3.Fol der part: The places where the resu mes are loaded and required skills according to the requirement of the job
where the output is stored. profile and this can be easy to find the correct fit for the
4.Output: The result of the application is stored in different required job.
forms based on the screening results.
5. Dataset: A CSV file which holds the skill set. NLP:
Natural language processing(see Figure.4) is the key
approach used in this web application. It is a subfield of
computer sciences and artificial intelligence concerned with
interactions between the machines and high level language.
In simple words it can be stated as how to program
applications to process and analyze large amounts of natural
language data.

Figure 3: Architechtural overview

978-1-6654-8328-5/22/$31.00 ©2022 IEEE 1774

Authorized licensed use limited to: ULAKBIM UASL - CUKUROVA UNIVERSITESI. Downloaded on November 09,2022 at 11:21:25 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Sixth International Conference on Trends in Electronics and Informatics (ICOEI 2022)
IEEE Xplore Part Number: CFP22J32-ART; ISBN: 978-1-6654-8328-5

Figure 4: Natural Language Processing

Here in this case we used NLP to scan the text in the
resumes by converting the pdf files into plain text files. To
do this we have used “en_core_web_sm”. Th is is used to
scan the plain text and to compare th is with the skills in skill Figure 6: Project Structure
dataset. Without using NLP, the process might be too •Model package consists of the various paths which makes
complex to find the frequency of skills in each resume. the engine local to any system in which it is running.
•Serv ice package consists of business logic which converts
MVC: the input files into pdfs and screens the files based on the
Model View Controller,as shown in Figure.5, is the design input dataset
•Controller package is the one which consists of controller
pattern which is usually followed by the programmers to
class and a sub package known as templates. The controller
make the code readable. According to this approach, the class is the one which handles the api requests . The
code is divided into layers instead of a single block. ‘temp lates’ is the package which responds on the api call
and displays the appropriate web pages[22].

VI. CODE W ALKT HROUGH

The application is basically written in python language

and various other techniques are used to make the application
programmer friendly and efficient. MVC design pattern is
followed in the application, such that the code is divided into
different parts based on their functionality.
The total code can be described in four layers,
Figure 5: Model View Controller
1. Controller Layer: It is the layer which consists of
The main advantage of this process is, each layer is loosely endpoint description and it is the one which d irectly
coupled to other layer. If any change is needed inside the communicates with the User Interface. The requests from
code, it can be done in a particular layer without disturbing user interface are directly directed to this layer.
other layer. In other case, if we write the applicat ion as a The requests from user are handled in the form of endpoints.
whole single block of code, any change in the code may Here in this case two endpoints are designed. “/check” to
break the entire flow. check and convert the files into pdf format and “/screen” to
perform the screening operation. The first endpoint provides
V. PROJECT ST RUCT URE the successful message after the execution. The second
endpoint provides JSON data of the resumes directly on the
The project structure of the proposed engine is designed as screen after the execution.
Model, View and Controller model that is shown in
2. Templates/View Layer: This is the layer in which
Figure.6. frontend code of the application is written. In a simple way
this layer can be stated as the combination of web pages.
This layer or package lies inside the controller package.
Here in this layer, mainly two buttons are designed and when
each button is clicked, the appropriate endpoint request is
sent to the controller.
3. Model Layer: This is the layer which usually consists of
the entities or identifiers of the application. In this case, the
paths of the folders which are handled by the application is
rested here. This makes the programmers to setup the
application in different systems easily by changing the paths
without disturbing other parts of the code.

978-1-6654-8328-5/22/$31.00 ©2022 IEEE 1775

Authorized licensed use limited to: ULAKBIM UASL - CUKUROVA UNIVERSITESI. Downloaded on November 09,2022 at 11:21:25 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Sixth International Conference on Trends in Electronics and Informatics (ICOEI 2022)
IEEE Xplore Part Number: CFP22J32-ART; ISBN: 978-1-6654-8328-5

4. Service Layer: This is the layer which consists of

business logic of the application. In this case this package
consists of two service classes. One of them holds the logic
to convert the files (doc/docx) into pdfs. The other class
holds the logic to screen the res umes based on the skill
dataset provided.
Upon “/check” request, the “ConvertToPdf.py” class is
triggered and checks the folder in which the resumes are
loaded. Then it counts the files with .doc and .docx
extensions and converts the files to pdfs. Upon “/screen”
request, the “Resume_service.py” class will start the
screening of the pdf files in the foler. After screening an
internal dataset was build based on the frequencies of the Figure 8: Screening flow
skills provided in the skill dataset. Finally the internal dataset Output: After the screening is performed based on the
will be plotted as a graph and stored in the output folder skillset, the output is stored in various forms like PNG, PDF
along with the csv file. Upon each api call success, a and CSV in the output folder(Figure.9 and Figure.10).
response is displayed on the screen. For ‘/check’ api, the
response string “SUCCESSFUL” is displayed and for
‘/screen’ api, the JSON data which is the result of the
screening will be rendered upon the successful screening of
resumes.

VII. W ORKING ST RUCT URE

The resumes which should be screened should be placed
in the input folder. The resumes can be in the word/pdf
format. When the code is compiled and gets run successfully
the screening process begins in the web. Initially the input Figure 9: Output Graph (PDF/PNG)
folder consists of the resumes which are to be screened. Later
after screening the input folder will be empty as the resumes
will be moved to destination. The total working consists of
two parts.
1. Coversion of resume files into pdf format
This flow will be triggered by hitting the URL “*/data”.
Three major formats like PDF, DOC, DOCX are accepted by
this application(see Figure.7). But before screening all the
docs and docx files are converted into pdfs for smooth flow.
Upon successful execution of this flow a message
“SUCCESSFUL” will be displayed on the screen. Figure 10: Output CSV

VIII. A DVANT AGES OF PROPOSED SOLUT ION

 Large number of resumes can be screened at a

time.
 Attained good time efficiency.
 Multiple formats like .pdf, .doc and .docx are
supported by this engine.
 The engine is designed in such a way that the input
Figure 7: Coversion to pdfs files(docx/doc/pdf) are converted into pdf’s and
2. Screening of Resumes then screened. This process makes the engine
accept multiple formats.
This is the main part of the application. This flow will be
 The NLP is used in such a way that total pdf is
triggered by hitting the URL “*/check”. The application
converted into plain text and then screened.
starts scanning the resumes in the input folder and generates
an internal dataset which further calculated and plotted as a  No filtering/deleting of words is involved, such that
bar graph against their name. Th is part of the application is there will be no extra t ime utilised to filer each
responsible for movement of files fro m input folder to resume.
destination folder. After Screening the output is stored inside  The data set provided to the engine is designed in
a folder named RESULT. Upon successful execution of this CSV format such that the addition of skills will be
flow the result JSON data is rendered and displayed on the easy to the user.
screen(see Figure.8).  The blocks of the code are designed in a loosely
coupled way, such that the flow will not be
disturbed when a new change is added.

978-1-6654-8328-5/22/$31.00 ©2022 IEEE 1776

Authorized licensed use limited to: ULAKBIM UASL - CUKUROVA UNIVERSITESI. Downloaded on November 09,2022 at 11:21:25 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Sixth International Conference on Trends in Electronics and Informatics (ICOEI 2022)
IEEE Xplore Part Number: CFP22J32-ART; ISBN: 978-1-6654-8328-5

 MVC design pattern is used to make the code [9] S. Himani, M. H. Kumar, M. K. Enduri, S. S. Begum, G. Rageswari
and S. Anamalamudi, "An Empirical Study on Impact of News
easily understandable. Articles," 2021 13th International Conference on Computational
 The output is stored in various formats like .png, Intelligence and Communication Networks (CICN), 2021, pp. 39-43,
.pdf and .csv for the convenience. doi: 10.1109/CICN51697.2021.9574670.
 The engine is designed in such a way that the files [10] R. Gupta, S. Sameer, H. Muppavarapu, M. K. Enduri and S.
Anamalamudi, "Sentiment Analysis on Zomato Reviews," 2021 13th
will automatically move to another folder after International Conference on Computational Intelligence and
processing. Communication Networks (CICN), 2021, pp. 34-38, doi:
 Some resumes which are not properly screened are 10.1109/CICN51697.2021.9574641.
moved in to a separate folder such that the user can [11] P. Gupta, A. Goswami, S. Koul and K. Sartape, "IQS-intelligent
take them in other way. querying system using natural language processing," 2017
International conference of Electronics, Communication and
 Simp le User Interface is designed to make it less Aerospace Technology (ICECA), 2017, pp. 410-413, doi:
complex for the user. 10.1109/ICECA.2017.8212846.
[12] A. Forestiero and G. Papuzzo, "Natural language processing approach
for distributed health data management," 2020 28th Euromicro
IX. CONCLUSION International Conference on Parallel, Distributed and Network-Based
Processing (PDP), 2020, pp. 360-363, doi:
Resume Screening is one of the most important steps in the 10.1109/PDP50117.2020.00061.
recruit ing process. Many ways are there to perform this [13] Y. Zhang, B. Bogard and C. Zhang, "Development of Natural
Language Processing Algorithm for Dental Charting," 2020 IEEE
process of screening resumes. Technology has made this 21st International Conference on Information Reuse and Integration
Resume Screen ing proces s easy for the recruiters. Even a for Data Science (IRI), 2020, pp. 403-404, doi:
large pool of applications can be screened easily and 10.1109/IRI49571.2020.00066.
efficiently with the help of technology. One of the best ways [14] K. Sintoris and K. Vergidis, "Extracting Business Process Models
discussed in this paper is the “Resume Screener”. It is very Using Natural Language Processing (NLP) Techniques," 2017 IEEE
19th Conference on Business Informatics (CBI), 2017, pp. 135 -139,
efficient and one of the best ways to screen a large pool o f doi: 10.1109/CBI.2017.41.
resumes at a time. “Resume Screener” is one of the best way [15] W. Maass, "How visual salience influences natural language
to be used in the resume screening process as it results in the descriptions," IEE Colloquium on Grounding Representations:
representation which can be very helpful in finding the best Integration of Sensory Information in Natural Language Processing,
fit for different requirements. Finally, “Resu me Screener” is Artificial Intelligence and Neural Networks, 1995, pp. 3/1-3/3, doi:
10.1049/ic:19950663.
a very useful application made using efficient technologies
[16] M. Ahirrao, Y. Joshi, A. Gandhe, S. Kotgire and R. G. Deshmukh,
like python and NLP, which results in saving time and "Phrase Composing Tool using Natural Language Processing," 2021
resources in the recruiting process for any job role. International Conference on Intelligent T echnologies (CONIT), 2021,
pp. 1-4, doi: 10.1109/CONIT51480.2021.9498546.
REFERENCES [17] Roberto Salazar, "Resume Screening with Python - Analyzing
Candidates Resumes for Jobs Openings", May 3, 2020
[1] Amit Singh, Catherine Rose, Karthik Visweswariah, Enara Vijil, [18] FreshWorks, “Resume screening - How to stop swimming in a sea of
Nandakishore Kambhatla, “PROSPECT : A system for screening resumes?”, 2016.
candidates for recruitment”, CIKM 2010, Toronto, Ontario, Canada, [19] S. Srivastava and S. Lamba, "Hindi-CNL Coder - A Desktop
October 26-30, 2010 Application for Learning Programming using Native Controlled
[2] Kajal Kumari, “ Resume Screening with Natural Language Processing Natural Language," 2020 IEEE 20th International Conference on
in Python”, June 23, 2021. Advanced Learning T echnologies (ICALT), 2020, pp. 228-229, doi:
10.1109/ICALT49669.2020.00074.
[3] R. Posevkin and I. Bessmertny, "Translation of natural language
queries to structured data sources," 2015 9th International Conference [20] Vamsi Naidu, P., Bommu, V.M.R., Pallapothu, V.D., Janapamula,
on Application of Information and Communication Technologies S.R.N., Kommuri, N.L. (2022). Resume Screening Using Machine
(AICT ), 2015, pp. 57-59, doi: 10.1109/ICAICT.2015.7338516. Learning. In: Chen, J.IZ., Tavares, J.M.R.S., Iliyasu, A.M., Du, KL.
(eds) Second International Conference on Image Processing and
[4] H. Isahara, "Resource-based Natural Language Processing," 2007 Capsule Networks. ICIPCN 2021. Lecture Notes in Networks and
International Conference on Natural Language Processing and Systems, vol 300. Springer, Cham.
Knowledge Engineering, 2007, pp. 11-12, doi:
10.1109/NLPKE.2007.4368002. [21] M. Bilenko, R. Mooney, W. Cohen, P. Ravikumar, and S. Fienberg.
Adaptive name matching in information integration. IEEE Intelligent
[5] C. T apsai, "Information Processing and Retrieval from CSV File by Systems, 18(5):16–23, 2003.
Natural Language," 2018 IEEE 3rd International Conference on
Communication and Information Systems (ICCIS), 2018, pp. 212 - [22] A. Z. Broder. Identifying and filtering near-duplicate documents. In
216, doi: 10.1109/ICOMIS.2018.8644947. COM ’00: Proceedings of the 11th Annual Symposium on
Combinatorial Pattern Matching, pages 1–10, London, UK, 2000.
[6] S. Weigelt and W. F. T ichy, "Poster: ProNat: An Agent-Based Springer-Verlag.
System Design for Programming in Spoken Natural Language," 2015
IEEE/ACM 37th IEEE International Conference on Software [23] Dharod, Grusha, Harshal Dedhia, Jaineel Shah, Smita Sankhe, and
Engineering, 2015, pp. 819-820, doi: 10.1109/ICSE.2015.264. Sneh Chitalia. "Trumer: Rumor Detection on T witter Using Machine
Learning and NLPTechniques." In Innovative Data Communication
[7] Pradeep KumarRoy, Sarabjeet SinghChowdhary, RockyBhatia, “A Technologies and Application, pp. 461-473. Springer, Singapore,
Machine Learning approach for automation of Resume 2021.
Recommendation system”,Procedia Computer Science 167 (2020)
2318–2327. [24] Dhawale, Apurva D., Sonali B. Kulkarni, and Vaishali Kumbhakarna.
"Survey of Progressive Era of T ext Summarization for Indian and
[8] A. Das and R. C. Balabantaray, "MyNLIDB: A Natural Language Foreign Languages Using Natural Language Processing." In
Interface to Database," 2019 International Conference on Information International Conference on Innovative Data Communication
Technology (ICIT), 2019, pp. 234-238, doi: T echnologies and Application, pp. 654-662. Springer, Cham, 2019.
10.1109/ICIT48102.2019.00048.

978-1-6654-8328-5/22/$31.00 ©2022 IEEE 1777

Authorized licensed use limited to: ULAKBIM UASL - CUKUROVA UNIVERSITESI. Downloaded on November 09,2022 at 11:21:25 UTC from IEEE Xplore. Restrictions apply.