Automated Resume Evaluation System using NLP

RohiniN imbekar1 , Y ogeshP atil2 , RahulP rabhu3 , ShainilaM ulla4

Department of Computer Engineering
Don Bosco Institute of Technology, Mumbai, India.
[email protected]

Abstract— Recruiting candidates to fit a particular job profile indicate that the recruiter takes just a minute to digest
is a task crucial to most of the companies. Due to increasing a resume completely. There are numerous sorts of CV
growth in online recruitment, traditional hiring methods are like combination resume written account resumes, targeted
becoming inefficient. The conventional techniques usually in-
clude a labor-intensive process of manually searching through resume, visual resumes, practical resumes, etc[1]
the applied candidates, reviewing their resumes, and then
producing a shortlist of suitable candidates to be interviewed. In this competitive world, a candidate ought to possess
In this era of technology, job searching has become smarter a strong resume that conveys the specified information
and more accessible at the same time. The companies receive required by a company for any specific designation. The
enormous numbers of resumes/CVs, which are not always
structured. There have been lots of work done for the job organization databases hold lakhs of resumes that are in
searching process. Whereas, the process of selecting a candidate free vogue and unstructured. The data and also the structure
based on their resume has not been entirely automated. This contents of resumes are going to be assortment below
research proposes a model of extracting valuable information sub topics; the classification and also the illustration of
from the resume and ranking it according to the preference data disagree from each other.[2] Thus, collection relevant
and requirement of the company. To achieve the desired goal,
the entire process has been divided into three segments. The information from every resume and storing it into the
first segment consists of converting the unstructured resumes database in a structured format would reduce human effort.
in structured data using NLP, and the second segment consists There are some difficulties in resume service by corporations
of the extraction phase, where the relevant information is since they consume abundant of your time, money, capacity,
extracted from the resume and giving them an identifier value. human effort. These companies need parsed resumes for
Finally, based on the values assigned,the resumes are ranked
accordingly in the final segment. achievement.

I. INTRODUCTION Automated recruitment systems need that Job seekers

While searching for jobs, the most important thing to post their resumes on numerous websites like LinkedIn,
represent an applicant is the resume. All the major industries Naukri.com, Monster.com, Resume builder, etc. Certain
today are driven by technology. The applicant can upload websites could retrieve unwanted resumes, whereas some
their resume. Resumes received are challenging to process may provide a terribly minimum range of resumes. This
and store in a unified database format. It becomes very involves an Associate in Nursing approach for qualitative
tedious to select the most appropriate ones. Since resumes analysis of resumes. Since there are numerous alternative
are unstructured documents based on the applicant’s writing existing websites provide advanced facilities like searching
skills, they can be created in a multitude of formats. Dynamic on the idea of keywords, domain, location, etc. Their search
extraction techniques are used to extract the most relevant doesn’t take into consideration, the skill level of a selected
information from the resumes. The filtering techniques match candidate.[3] For example, if an organization searches for
the resumes from the database to a single job posting. Re- a willing candidate who can work in python language,
sumes obtained by these filters are generally similar to each they will look for candidates who have python language
other as they fulfil the same criteria posted by the employer. mentioned in their resumes. But how will they recognize the
This system fetches resumes which satisfy the requirement proficiency of that individual candidate in python language?
of a particular job post and rank them accordingly.
In our epitome, we will include other parameters like
II. RELATED WORK the projects and other skills like learning software in which
The hiring process has been changing gradually over the the willing candidate can embrace in their resume. This
amount of your time. Graduates aim for getting placed in information is going to be considered as input from the
a good company whereas the recruiters search for people candidate, and by analyzing the text employed by him
who would contribute effectively for structure growth. or her, we have a tend to will categorize the candidate
The primary impression of each candidate is their resume. into numerous expertise levels. So if a company need
Resume structure and its content are live for the leader an associate employee who mainly incorporates a high
to form a call for holding or eliminating the candidate experience level in python language, solely then his resume
for future rounds of any achievement method. Statistics can be shortlisted. A knowledge domain of different

Authorized licensed use limited to: PES University Bengaluru. Downloaded on January 12,2025 at 13:21:15 UTC from IEEE Xplore. Restrictions apply.
keywords can be designed, which can type the idea of B. Selection
categorization. This will conjointly facilitate the ranking, of Selection may be a later stage of achievement. It is the
multiple resumes to inform which one is better or worse method of choosing people United Nations agency have
than the opposite allowing the apply candidate to present relevant qualifications to fill jobs in a company. Choice is
themselves within the best possible method. The section way quite simply selecting the simplest candidate. It’s a trial
identification and pdf parsing can also be improved. The to strike a balance between what the organization need and
standard reference documents can be augmented. The what candidate need to try. It involves selecting not solely
algorithmic formulation has the possibility of being used in new members of the organization however, additionally
an annotation and recommendation system.[4] guaranteeing that the choice method will manage to attract
competent and qualified candidates suited to work.
III. T HE EXISTING RECRUITMENT PROCESS The focus in the selection process is on:
(1) The method of selection and skills in terms of
Traditionally achievement and selection are two necessary contribution to the dependability of selections.
methods of human resource management. The purpose of (2) The factor outlined and applied by decision-makers. And
achievement is to spot appropriate manpower to fulfill the the way these reflect their comprehension of “necessary
work necessities and job specifications. It is the foremost competence”.
necessary perform of personnel administration on the op- (3) However, the processes of selection embody the
posite hand choice is bothered with securing the right info assumptions and commitments, the generalities, truths and
regarding someone. The article of choice method is to confusions—of decision-makers about the imperatives of
see whether or not the prospective candidate possesses the organization’s culture and also the means they request to
qualification for the particular job it’s, an extended method. require care of and change this.
It Starts from the interview and ends with the contract of
employment.[5]

A. Recruitment
HR plays essential role in achieving all the requirements
of every department of a company. The strategic desires of
the organization are specially designed. The activity that
comes below the achievement method is to rent a worker
for a corporation to confirm the continued operation of the
organization.[6] This involves act with actual or potential
job seekers, attract them to own a chance, and try to
persuade them to figure for the organization in any means
that of the fields. The target to attain the simplest quality
result’s to own the simplest quality within the best manner
and numbers.

In the past, CVs/Resumes submitted by job seekers

accustomed to be manually analyzed and judged by the
employers [3]. This methodology remains followed within
recent times. There is a range of various sources to recruit
personnel reckoning on the sort of job vacancy. There are Fig. 1. Traditional recruitment process
numerous companies that have totally different necessities
for his or her organization. However, a massive-scale
company receives many resumes every day, thus handling IV. RESEARCH METHODOLOGY
those numbers of resumes has become a crucial task
and time overwhelming method. Due to these reasons, Based on previous literature review form was developed
numerous companies provide specific format for job seeker. that unstructured data contain virtually real-world data are
Jobseeker ought to replenish with needed info then the difficult to classify since there’s no predefined structure.
CV/Resume is going to be analyzed by machine, with easy
pattern recognition and keywords looking. Whereas this A. Conversion Phase
methodology reduced the employment for the employers, The first process that comes into existence the conver-
it exaggerated the quantity of labor for the candidates sion of unstructured data to structured data. It has three
considerably as they have to maintain different formats for main stages delineated in figure 2. Database consists of
every job they apply. To boot, it additionally tends to scale unstructured data, data in its negotiant state and successfully
back the power and also the flexibility of writing the abilities structured data. The user uploads the resumes. This can be
at the side of the qualifications during a CV/Resume.[3] automatically stored in a database.

Authorized licensed use limited to: PES University Bengaluru. Downloaded on January 12,2025 at 13:21:15 UTC from IEEE Xplore. Restrictions apply.
 D. Ranking

Ranking the resume can facilitate the organization to

choose the candidate wisely. The ranking process will in
shortlisting variants of CV. The increasing number of submit-
ted CVs could overwhelm HR departments, which usually
perform manual analysis of job applications. This process
will improve the decision-making process, careful parame-
terization by the department’s skilled recruiters which helps
in better recruitment process in a shorter period.
Our system uses a machine learning algorithm to build
Fig. 2. Conversion of unstructured data to structured data the applicant learning model. It requires training data to
learn the algorithm for constructing a ranking model, so it
will predict the recruiter’s judgment once given the candi-
B. Extraction Phase date’s resume. To combine predefined features for ranking,
supervised learning algorithms are called “learning-to-rank”
Resumes are dissimilar and have different structure. The methods is used.[9]
resume consists of many information like education, skills,
personal information, achievements, internships, etc. All this In the testing phase, the candidate application is applied
information in a resume varies considerably. Our proposed to learn model for sorting, and then finalized rank candidates
system will work on unstructured resume. It will extract list is generated.
unstructured resumes.[9]
Entity Name Extraction (NER). It helps in the transfor-
mation of unstructured text to structured text. Input resume
will be in any format like web posts, PDFs, documents,
spreadsheets, etc. NER is used for recognizing the entities
into different parameters such as place, people, numerical
expressions like phone number, time, etc. likewise temporal
expression like durations, frequency, etc. It is a subtask of the
information extraction method. Once a CV is structured, then
extracting the relevant information becomes a straighforward
task.[8]

Fig. 4. Ranking Process

V. A RCHITECTURE OVERVIEW

The proposed system consists of multiple modules. First

the Section-Based Segmentation module which is used to
extract the candidate’s information like personal, education,
experience, technical skills, internship(if any), hobbies, etc.
Fig. 3. Extraction Process
The next module is the filtration module, which refines the
lists by removing the insignificant terms that do not con-
tribute to the matching process[10]. The third module takes
C. Filtration
a set of skills extracted from both resumes and job portals as
The process of filtering resumes is predicated on compar- input to classify them under their corresponding occupational
ing the job skills posted by the employer with the extracted categories. The category-based matching module takes the
data from the resumes. This process gives all the applicants list of skills from both the resume and job posts to construct
who match the job description. To make the filtration process the semantic network. By deriving the relatedness between
more efficient, a score is given to each resume to rank the their concepts. Finally, the matching algorithm takes the
applicant. Collaborative filtering is used to to predict the semantic network as input and produces the measures of
trend of selection [11]. closeness between them as output.

Authorized licensed use limited to: PES University Bengaluru. Downloaded on January 12,2025 at 13:21:15 UTC from IEEE Xplore. Restrictions apply.
 resume may differ from company to company. As described,
the whole process was segmented, and each segment was
designed separately to perform its task. Finally, ARES gives
ranks to resumes based on the necessary data, and the
employers take the top few applicants into consideration.
VIII. FUTURE SCOPE
In this research, the domain was restricted to the resumes
of only engineering students, and the amount of sample
data versus the amount of test data was relatively small.
The domain can be extended further to other domains like
Fig. 5. Architecture Overview Telecommunication, Healthcare, E-commerce, and public
sector jobs. In the future, we also plan to utilize the extracted
information applicants’ resumes to dynamically generate user
VI. SURVEY REVIEW profiles to be further used for recommending a job to job
seekers. Also, this system can be extended by including
Sr No. Title Work done by Gap Identified aptitude tests. Therefore, future scope is vast.
JRC A Job
R EFERENCES
Post and
Abeer Zaroor, The resumes [1] Ryland, E. K., amp; Rosen, B. (1987). Personnel Professionals Re-
Resume
Mohammed used in this actions to Chronological and Functional Résumé Formats [Abstract].
1. Classification The Career Development Quarterly, 35(3), 228- 238.
Maree Muath research was
System [2] R. Fagin, B. Kimelfeld, Y. Li, et al. Understanding queries in a search
Sabha structured database system. In PODS, pages 273–284, 2010
for Online
[3] Malamitsa, K., Kokkotas, P., amp; Kasoctas, M. (december 2008).
Recruitment Graph/Chart Interpretation and Reading Comprehension as Critical
Thomas Thinking Skills. Science Education International, 19(4)
Matching Jobs The resumes [4] Di Wu, Lanlan Wu, Tieli Sun, Yingjie Jiang, “Ontology based infor-
Schmitt,
and Resumes: tokens were mation extraction technology”, International Conference on Internet
Philippe Technology and Applications (iTAP), Aug 2011.
2. a Deep Col- extracted in
Caillou, and [5] Essential of human resource management and industrial relations,
laborative Fil- structured subbarao Mumbai, Himalaya publishing house, 1996
Michele
tering Task format [6] Abbie, R., Robert, V. (2008). Moving beyond traditional recruitment
Sebag strategies. Recruiting and Hiring—Research Paper. Retrieved from
Automatic http://academics.lmu.edu/media/lmuacademics/diversity/images/Moving
Beyond Traditional Recruitment Strategies.pdf
Text [7] “Entity extraction: How does it work? - expert system,” www.
Youngjoong
Categorization The categories expertsystem.com/entity-extraction-work/.
3. Ko, Jungyun [8] “Named entity extraction—lexalytics,” https://www.lexalytics.com/
by are predefined.
Seo technology/entity-extraction.
Unsupervised [9] Finn, A. and Kushmerick, N. ”Multi-level boundary classification for
Learning infonnation extraction,” in Proceedings of 15th European Conference
Application Evanthia on Machine Learning, Pisa, Italy, September 20-24, 2004.
[10] A Kmail, M Maree, M Belkhatir, and S Alhashmi ”An Automatic
of Machine Faliagka, Inconsistent Online Recruitment System based on Exploiting Multiple Semantic
Learning Kostas CV for- Resources and Concept-relatedness Measures,” Proceedings of the
4. Algorithms Ramantas, mats,structured EEE 27th International Conference on Tools with Artificial Intelli-
gence (ICTAI), pp. 620-627, 2015
to an online Athanasios and contextual [11] Chanawee Chanavaltada, Panpaporn Likitphanitkul, Manop
Recruitment Tsakalidis information. Phankokkraud, ”An Improvement of Recommender System to
System Find Appropriate Candidate for Recruitment with Collaborative
Filtering”, published in 2015 ICCSS
Effective
Cluster-based
ranking can
Ranking Index
improved
for Enhancing
through
Recruitment Mayuri
5. supervised
Process using Verma
machine
Text Mining
learning
and Machine
applied in past
Learning
dataset.
VII. CONCLUSION
The proposed model in this paper extracts the necessary
data from a resume and segments them based on their values.
However, the ranking and weights given for a segment of the

Authorized licensed use limited to: PES University Bengaluru. Downloaded on January 12,2025 at 13:21:15 UTC from IEEE Xplore. Restrictions apply.