SCREENING AND RANKING RESUME USING STACKED MODEL

Submitted in partial fulfillment of the

requirements for the award of
Bachelor of engineering degree in Computer Science and Engineering

By

ABIRAAMI S (Reg. No - 391100012)

ABINAYA R (Reg. No - 39110010)

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

SCHOOL OF COMPUTING

SATHYABAMA
INSTITUTE OF SCIENCE AND TECHNOLOGY
(DEEMED TO BE UNIVERSITY)
Accredited with Grade “A” by NAAC | 12B Status by UGC | Approved by AICTE
JEPPIAAR NAGAR, RAJIV GANDHISALAI,
CHENNAI - 600119

APRIL - 2023
 DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

BONAFIDE CERTIFICATE

This is to certify that this Project Report is the bonafide work of ABIRAAMI.S
(39110012) and ABINAYA.R (39110010) who carried out the Project Phase-2
entitled “SCREENING AND RANKING RESUME USING STACKED MODEL”
under my supervision from January 2023 to April 2023.

Internal Guide
Dr. S. PRINCE MARY, M.E., Ph.D.

Head of the Department

Dr. L. LAKSHMANAN, M.E., Ph.D.

Submitted for Viva voce Examination held on 20.04.2023

Internal Examiner External Examiner

 DECLARATION
ii

I, ABIRAAMI.S (Reg. No: 39110012), hereby declare that the Project Phase-2
Report entitled “SCREENING AND RANKING RESUME USING STACKED
MODEL” done by me under the guidance of Dr. S. Prince Mary, M.E., Ph.D.
is submitted in partial fulfillment of the requirements for the award of Bachelor
of Engineering degree in Computer Science and Engineering.

DATE:20. 04.2023
ABIRAAMI.S
PLACE: Chennai SIGNATURE OF THE CANDIDATE

iii
 ACKNOWLEDGEMENT

I am pleased to acknowledge my sincere thanks to Board of Management of

SATHYABAMA for their kind encouragement in doing this project and for
completing it successfully. I am grateful to them.

I convey my thanks to Dr. T. Sasikala M.E., Ph. D, Dean, School of Computing,

Dr. L. Lakshmanan M.E., Ph.D., Head of the Department of Computer Science
and Engineering for providing me necessary support and details at the right time
during the progressive reviews.

I would like to express my sincere and deep sense of gratitude to my Project

Guide Dr. S. Prince Mary, M.E., Ph.D. for her valuable guidance, suggestions and
constant encouragement paved way for the successful completion of my phase-1
project work.

I wish to express my thanks toall Teaching and Non-teaching staff members of the
Department of Computer Science and Engineering who were helpful in many
ways for the completion of the project.

iv
 ABSTRACT

The evolving technology is creating many chances of employment for

many. Nowadays to apply for any job the most essential document is a resume. A
resume tells a lot about the person's achievements and the skill sets in all walks of
life. The person applying for the job highlights the strong points and skill sets
required for the company. Multinational organizations receive thousands of emails
from such people who send their resumes for them to apply for a certain post. Now
the real challenge is to know which resume is to be sorted and shortlisted
according to the constraints. One method is to manually check and sort the
resume. Now, this method is the most time-consuming and also can lead to a lot of
errors because of human interventions. Also, humans cannot keep on working
continuously. Using Natural Language Processing and Machine Learning to rank
the resumes according to the given constraint, this intelligent system ranks the
resume of any format according to the given constraints or the following
requirement provided by the client company. We will basically take the bulk of
input resume from the client company and that client company will also provide the
requirement and the constraints according to which the resume should be ranked
by our system.

v
 TABLE OF CONTENTS
Chapter Page
TITLE
No. No.
ABSTRACT v

LIST OF FIGURES viii

LIST OF ABBREVIATIONS ix

1 INTRODUCTION 1

2 LITERATURE SURVEY 4

2.1 Inferences from Literature Survey 7

2.2 Open problems in Existing System 7

2.3 Scope of the project 7

3 REQUIREMENTS ANALYSIS 8

3.1 Feasibility Studies 8

3.2 System Use case 10

3.3 Software Requirement Specification 11

4 DESCRIPTION OF PROPOSED SYSTEM 12

4.1 Selected Methodology or process model 12

4.2 Architecture / Overall Design of Proposed System 13

Description of Software for Implementation and 14
4.3
Testing plan of the Proposed Model/System
4.4 Project Management Plan 18
5 IMPLEMENTATION DETAILS 19

5.1 Algorithms 19

5.2 Existing System 24

5.3 Proposed System 24

6 RESULTS AND DISCUSSION 27
7 SUMMARY 29

7.1 Conclusion 29

7.2 Future work 30

REFERENCES 31

vi
APPENDIX 33
A SOURCE CODE 33
B SCREENSHOTS 46
C RESEARCH PAPER 49

vii
 LIST OF FIGURES

FIGURE NO. FIGURE NAME PAGE NO.

3.2 USECASE DIAGRAM 10

4.2 ARCHITECTURE DIAGRAM 14

4.4 PROJECT MANAGEMENT PLAN 18

6.1 KNEIGHBOURS CLASSIFIER 27

6.2 CONFUSION MATRIX 27

6.3 WORD CLOUD 28

viii
 CHAPTER 1

INTRODUCTION

Day by day there are so many new researches that are being carried out
by so many organizations in many fields. Right from the IT sector to the Medical
fields, there are so many researches as well are progress made in day to day
life. These progress in these fields are creating new employment opportunities all
over the world. Now we know that the hiring process across the world is the
same. That is the candidate has to make the resume first and as most of the
recruitment process is based on the candidates' resume. Now most of the
organizations tell the candidates to send the resume via e-mails. Now after they
receive the email the next job is to sort them according to the requirement. Now
usually they sort these manually but this process is time consuming.

Also there may be so many errors due to that. The efficiency of such
sorting is extremely low. This may result in giving a chance to a candidate whose
credentials are not up to the mark for the organization or this may also make the
manual sorting miss the candidate who is extremely good for the organization.
Hence a system which is intelligent enough to sort all these documents without
any error and within time is needed which is exactly our aim. Machine learning
for recruiting is the application of machine learning, such as the learning or
problem-solving that a computer can do, to the recruitment function. This new
technology is designed to streamline or automate some part of the recruiting
workflow, especially repetitive, high-volume tasks. For example, software that
applies machine learning to resumes to auto-screen candidates or software that
conducts sentiment analysis on job descriptions to identify potentially biased
language. The current recruitment process is more tedious and time consuming
which forces the candidates to fill all their skill and information manually. And HR
team requires more man power to scrutinize the resumes of the candidates. So
that motivated to build a solution that is more flexible and automated.

1
 Recruitment is a 200-billion-dollar business. It deals with hiring the best-fit
candidates having the relevant skills for a given job profile from an immensely
large pool of candidates. If a company has any job opening for a position, scores
of candidates mail their resumes to the company to apply for that opening. In the
hiring process, the first task for any recruiter is to screen the resumes of all the
job applicants. Any company having a job opening for a particular position will
have their mail inboxes bombarded with thousands of emails from aspiring job
applicants every single day. Selecting the prospective candidates for that job
position from a large pool of candidates for any recruiter is very tedious. It is an
extremely daunting task for the recruiters of a company to manually go through
thousands of resumes and select the most appropriate candidates for the job.
Out of those thousands of resumes submitted to the company for the given job
posting, about 75% of them do not showcase the relevant skills that are required
for the job profile.

Due to this, the recruiters quite often find it really arduous to narrow down
the most appropriate candidates from a large applicant pool. In recent years,
there have been more than 50,000 e-recruitment sites have been developed.
The developers of these online recruitment sites have used various approaches
to identify the prospective candidates for a given job profile of a company. Some
of these, have managed to employ classification techniques that will classify the
candidate resumes into various categories for every job posting given by every
company. In these approaches, every candidate’s resume is tried to match with
every given job posting on the recruitment site. The aim of these recruitment
sites is to throw up the results to the candidate to which they are the best fit. The
techniques used by these sites have resulted in high accuracy and precision, but
one of the major disadvantages is the factor of time complexity. If every
candidate’s resume is matched with every other job posting given on the online
recruitment site, the time complexity for acquiring the results is very high.

The world of Artificial Intelligence [AI] and Machine Learning [ML] has
grown significantly in recent times. The availability of large amounts of data

2
brought about by advancements in technology which has made the internet
cheap and accessible to previously inaccessible regions of the world has
contributed to a great increase in the performance of the ML models in recent
times. Software companies around the world exploit the advances in ML to drive
automation and increase their productivity in areas that relied mostly on manual
human labor. The approach discussed in this project is by using machine
learning to train the dataset for a particular type of job position. It is also
proposed to use section-based segmentation for data extraction using Natural
language Processing (NLP).

In order to improve the time efficiency of the web application, the

candidate’s resume will only be matched to those job openings where they are
interested in and have applied to which will, in turn, reduce the time complexity.
Besides, the results of the resume matching of all the candidates who have
applied for the job opening will be visible only to the recruiter of that particular
company. This is done with the aim to aid the recruiters of any company from the
long and tedious task of viewing and analyzing thousands of candidates’
resumes. In this intelligent-based approach, they will be given the option to view
the candidate’s resume as well as they will get the results of the best candidates
suitable for the required job position. Same disease, we use it as the score to
rank the similarity among patients

3
 CHAPTER 2
LITERATURE SURVEY

This problem statement has been extensively studied over the past 5
years by researchers and automotive companies in a bid to create a solution,
and all their solutions vary from analyzing various patterns of distractive habits to
analyzing resume.

Abdul Samad Ebrahim Yahiya [1] this resumed striving to reach a review
of the tools and the necessary methodology to present a clear understanding of
the association of NLP and DL for truly understanding in the training. Efficiency
and execution both are improved in NLP by Part of speech tagging (POST),
Morphological Analysis, Named Entity Recognition (NER), and Semantic Role
Labeling (SRL), Syntactic Parsing, and Conference resolution. Artificial Neural
Networks (ANN), Recurrent Neural Network (RNN), Convolution Neural
Networks (CNN), dealings among Dense Vector (DV), Windows Approach (WA),
and Multitask learning (MTL) as a characteristic of Deep Learning.

Anushka Sharma[2] , the research work in [7] focuses on extracting data

from resumes and performing the required analysis on the data to convert it into
useful information for the recruiters. Thus, the Resume Parser would help the
recruiters to select the best relevant candidates in a minimal amount of time,
consequently saving their time and effort.

Elizabeth George [3] suggested that, the proposed system, JARO

accelerates the interview process towards an unbiased decision-making process
by proposing a chatbot that would conduct interviews by analyzing the
candidate’s Curriculum Vitae (CV), based on which, it then prepares a set of
questions to be asked to the candidate. The system will consist of features like
resume analysis and automatic interview processes.

4
 Garima Bhardwaj [4] The study work in [8] was conducted among 115 HR
professionals at various IT sectors in Delhi/NCR region. A multiple regression
method was used to test the hypothesis and confirmed a positive relationship
between these two factors establishing about the increased use of AI at work
results in better HR functional performance. However, AI has a significant
relationship with innovativeness and also with the ease of use which reflects AI
affects HR with innovations and ease of use.

Huaizheng Zhang [5], the author in [8] developed a method for automatic
RQA. Since there is also no public dataset for model training and evaluation, we
build a dataset for RQA by collecting around 10K resumes, which are provided
by a private resume management company. By investigating the dataset, we
identify some factors or features that could be useful to discriminate good
resumes from bad ones, e.g., the consistency between different parts of a
resume.

Muath Sabha [6] the research work of author in [5] presented a hybrid
approach that employs conceptual-based classification of resumes and job
postings and automatically ranks candidate resumes (that fall under each
category) to their corresponding job offers. In this context, the exploit an
integrated knowledge base for carrying out the classification task and
experimentally demonstrate - using a real-world recruitment dataset- achieving
promising precision results compared to conventional machine learning-based
resume classification approaches

Mujtaba [7] presents an overview of fairness definitions, methods, and

tools as they relate to recruitment and establishes ethical considerations in the
use of machine learning in the hiring space. Considering Deep Learning (DL)
method recognize artificial Neural Network (NN) to nonlinear process, NLP tools
become increasingly accurate and efficient that begins a debacle. Multi-Layer
Neural Network obtaining the importance of the NLP for its capability including
standard speed and resolute output. Hierarchical designs of data operate
recurring processing layers to learn and with this arrangement of DL methods
manage several practices.

5
 Pradeep Kumar Roy [8] suggested that an automated way of “Resume
Classification and Matching” could really ease the tedious process of fair
screening and shortlisting; it would certainly expedite the candidate selection and
decision making process. This system could work with a large number of
resumes for first classifying the right categories using different classifier, once
classification has been done then as per the job description, top candidates
could be ranked using Content-based Recommendation, using cosine similarity
and by using k-NN to identify the CVs that are nearest to the provided job
description.

Sanjay Revanna [9] proposed the top applicants might be rated using
content-based suggestion, which uses cosine similarity to find the curriculum
vitae that are the most comparable to the job description supplied and KNN
algorithm is used to pick and rank Curriculum Vitaes (CV) based on job
descriptions in huge quantities.

The work of Sharadadevi Kaganurmath [10] created an automated machine

learning-based algorithm that recommends acceptable applicant resumes to HR
based on the job description provided. The suggested methodology had two
stages: first, it classified resumes into various groups. Second, it suggests
resumes based on their resemblance to the job description. If an industry
produces a high number of resumes, the proposed approach can be used to
create an industry-specific model.

6
2.1 INFERENCES FROM THE LITERATURE SURVEY

Resume Screening is a very important step in the hiring process. Even

today, many companies are following the traditional resume screening process
Even though many efficient systems were proposed to perform resume
screening operations, due to some problems; the systems are not efficient as
expected. Resume Screening using Machine Learning approaches has some
issues like a time consuming, not being user-friendly, etc. Resume Screening
using Artificial Intelligence approaches has some issues with input/output format,
single resume acceptance for a time (which leads to a lot of time to screen all the
resumes), etc.

2.2 OPEN PROBLEMS IN EXISTING SYSTEM

The systems has models that don’t have any way to improve themselves
over the time, the models will be trained only once. The models used Machine
learning algorithms which have a tend to plateau in performance when run over a
large dataset. The application can only be used effectively by the programmer.
Each code block is interdependent; each change requires disrupting the entire
code, which may disrupt the flow.

2.3 SCOPE OF THE PROJECT

Resume screening is the process of quickly reviewing a resume while
looking for keywords and characteristics. It saves recruiters time during the hiring
process and leads to faster onboarding of new team members. Whether you
screen resumes manually or with software, using the best techniques can help
you identify the most suitable candidates. it is the process of quickly reviewing a
resume while looking for keywords and characteristics. It saves recruiters time
during the hiring process and leads to faster onboarding of new team members.
Whether you screen resumes manually or with software, using the best
techniques can help you identify the most suitable candidates.

7
 CHAPTER 3

REQUIREMENTS ANALYSIS

Software requirement means requirement that is needed by software to

increase quality of software product. These requirements are generally a type of
expectation of user from software product that is important and need to be fulfilled
by software. Analysis means to examine something in an organized and specific
manner to know complete details about it. Therefore, Software requirement
analysis simply means complete study, analyzing, describing software
requirements so that requirements that are genuine and needed can be fulfilled to
solve problem. There are several activities involved in analyzing Software
requirements.

3.1 FEASIBILITY STUDIES

Recruitment is a 200-billion-dollar business. It deals with hiring the best-fit

candidates having the relevant skills for a given job profile from an immensely
large pool of candidates. If a company has any job opening for a position, scores
of candidates mail their resumes to the company to apply for that opening. In the
hiring process, the first task for any recruiter is to screen the resumes of all the job
applicants. Any company having a job opening for a particular position will have
their mail inboxes bombarded with thousands of emails from aspiring job
applicants every single day. Selecting the prospective candidates for that job
position from a large pool of candidates for any recruiter is very tedious. It is an
extremely daunting task for the recruiters of a company to manually go through
thousands of resumes and select the most appropriate candidates for the job. Out
of those thousands of resumes submitted to the company for the given job posting,
about 75% of them do not showcase the relevant skills that are required for the job
profile.

Due to this, the recruiters quite often find it really arduous to narrow down
the most appropriate candidates from a large applicant pool. In recent years, there
have been more than 50,000 e-recruitment sites have been developed. The

8
developers of these online recruitment sites have used various approaches to
identify the prospective candidates for a given job profile of a company. Some of
these, have managed to employ classification techniques that will classify the
candidate resumes into various categories for every job posting given by every
company. In these approaches, every candidate’s resume is tried to match with
every given job posting on the recruitment site. The aim of these recruitment sites
is to throw up the results to the candidate to which they are the best fit. The
techniques used by these sites have resulted in high accuracy and precision, but
one of the major disadvantages is the factor of time complexity. If every
candidate’s resume is matched with every other job posting given on the online
recruitment site, the time complexity for acquiring the results is very high.

The world of Artificial Intelligence [AI] and Machine Learning [ML] has
grown significantly in recent times. The availability of large amounts of data
brought about by advancements in technology which has made the internet cheap
and accessible to previously inaccessible regions of the world has contributed to a
great increase in the performance of the ML models in recent times. Software
companies around the world exploit the advances in ML to drive automation and
increase their productivity in areas that relied mostly on manual human labor. The
approach discussed in this project is by using machine learning to train the dataset
for a particular type of job position. It is also proposed to use section-based
segmentation for data extraction using Natural language Processing (NLP). In
order to improve the time efficiency of the web application, the candidate’s resume
will only be matched to those job openings where they are interested in and have
applied to which will, in turn, reduce the time complexity. Besides, the results of
the resume matching of all the candidates who have applied for the job opening
will be visible only to the recruiter of that particular company. This is done with the
aim to aid the recruiters of any company from the long and tedious task of viewing
and analyzing thousands of candidates’ resumes. In this intelligent-based
approach, they will be given the option to view the candidate’s resume as well as
they will get the results of the best candidates suitable for the required job position.
Same disease, we use it as the score to rank the similarity among patients.

9
3.2 SYSTEM USE CASE

This Use Case Diagram is a graphic depiction of the interactions among the
elements of Resume Builder. It represents the methodology used in system
analysis to identify, clarify, and organize system requirements of Resume Builder.
The main actors of Resume Builder in this Use Case Diagram are: Dataset, User,
Algorithm who perform the different type of use cases such as Manage Resume,
Manage Skills, Manage Job, Manage Formats, Manage Resume Types, Manage
Users and Full Resume Builder Operations. Major elements of the UML use case
diagram of Resume Builder are shown on the Fig 3.2.

Fig 3.2 – Use-Case Diagram

10
3.3 SOFTWARE REQUIREMENTS SPECIFICATION DOCUMENT

3.3.1 HARDWARE REQUIREMENTS

Processor : Pentium Dual Core 2.00GHZ

Hard disk : 120 GB
RAM : 2GB (minimum)
Keyboard : 110 keys enhanced

3.3.2 SOFTWARE REQUIREMENTS

Operating system : Windows7 (with service pack 1), 8, 8.1 and 10

IDE : Anaconda
Language : Python

11
 CHAPTER 4

DESCRIPTION OF PROPOSED SYSTEM

This system is a resume ranking software that uses natural language

processing (NLP) and machine learning. Following resume screening, the software
rates prospects in real-time depending on the recruiter's job needs. The web
application aims to order the resumes, by intelligently reading job descriptions as
input and comparing the resumes which fall into the category of given Job
Descriptions. It provides a ranking after filtering and recommends the better
resume for a given textual job description. In order to match and rate candidates in
real-time, the software employs natural language processing.
This work takes a different approach as it focuses mainly on the content of
the resumes where the extraction of skills are performed and related parameters
to match candidates with the job descriptions

4.1 SELECTED METHODOLOGY OR PROCESS MODEL

4.1.1 NATURAL LANGUAGE PROCESSING (NLP)

Resume collection is being performed and folder Structure creation is being

done. Data Cleaning such as removing clutter and unnecessary punctuation would
be taken off, Feature Engineering would be performed for enhancement This
includes removing stop words, punctuation, and stemming. This process will
construct a graph with sentences as the vertices. Importing necessary libraries is
performed. This library creates a summary of the supplied information within the
word limit. With the help of Natural language processing techniques, the
application which extracts the names of people, places, and other entities from
text, the main goal being to get a reduced version of it that retains the most
important information. It is followed by Exploratory Data Analysis. Exploratory data
analysis (EDA) is often a necessary task in uncovering hidden patterns, detecting
outliers, and identifying important variables and any anomalies in data.

12
4.1.2 TF-IDF

At this stage, a dynamic Script for the Tf-Idf approach is written. Term
frequency-inverse document frequency, is a numerical statistic that is intended to
reflect how important a word is to a document in a collection.TF-IDF is word
frequency scores that aim to emphasize phrases that are more interesting, e.g.,
common in a text but not across texts, without delving into the arithmetic. The TF-
IDF Vectorizer tokenizes texts, learns vocabulary, inverts frequency weightings,
and allows encoding new ones. It provides information on a word frequency in the
documents. Higher the TF- IDF score of a term which is computed using the above
equations represents more relevance in a document.

4.1.3 LATENT DIRICHLET ALLOCATION (LDA)

A tool and technique for Topic Modeling, Latent Dirichlet Allocation (LDA)
classifies or categorizes the text into a document and the words per topic, these
are modeled based on the Dirichlet distributions and processes. Latent Dirichlet
Allocation has been used in the application for the following functions- Discovering
the hidden themes in the data. Classifying the data into the discovered themes.
Using the classification to organize/summarize/search the documents. The
application, then deals with the calculation of the score for a candidate’s resume
according to the job posting they have applied for. According to the score each
candidate’s resume receives, a rank list will be made with the candidate receiving
a higher score placed higher as compared to the candidate receiving a lower
score.

4.2 ARCHITECTURE OF PROPOSED SYSTEM

This diagram is nothing but a simple description of all the entities that have
been incorporated into the system. The diagram represents the relations between
each of them and involves a sequence of decision-making processes and steps.
You can simply call it a visual or the whole process and its implementation. All
functional correspondences are explained in this Fig 4.2.

13
 Fig 4.2 – Architecture Diagram

4.3 DESCRIPTION OF SOFTWARE FOR IMPLEMENTATION AND

TESTING PLAN OF THE PROPOSED MODEL

4.3.1 PLATFORM SPECIFICATION– ANACONDA

Anaconda is an open-source package manager for Python and R. It is the

most popular platform among data science professionals for running Python and R
implementations. There are over 300 libraries in data science, so having a robust
distribution system for them is a must for any professional in this field. Anaconda
simplifies package deployment and management. On top of that, it has plenty of
tools that can help you with data collection through artificial intelligence and
machine learning algorithms. With Anaconda, you can easily set up, manage, and
share Conda environments. Moreover, you can deploy any required project with a
few clicks when you’re using Anaconda. There are many advantages to using
Anaconda and the following are the most prominent ones among them: Anaconda

14
is free and open-source. This means you can use it without spending any money.
In the data science sector, Anaconda is an industry staple. It is open-source too,
which has made it widely popular. If you want to become a data science
professional, you must know how to use Anaconda for Python because every
recruiter expects you to have this skill. It is a must-have for data science.

It has more than 1500 Python and R data science packages, so you don’t
face any compatibility issues while collaborating with others. For example,
suppose your colleague sends you a project which requires packages called A and
B but you only have package A. Without having package B, you wouldn’t be able
to run the project. Anaconda mitigates the chances of such errors. You can easily
collaborate on projects without worrying about any compatibility issues. It gives
you a seamless environment that simplifies deploying projects. You can deploy
any project with just a few clicks and commands while managing the rest.
Anaconda has a thriving community of data scientists and machine learning
professionals who use it regularly. If you encounter an issue, chances are, the
community has already answered the same. On the other hand, you can also ask
people in the community about the issues you face there, it’s a very helpful
community ready to help new learners. With Anaconda, you can easily create and
train machine learning and deep learning models as it works well with popular
tools including TensorFlow, Scikit-Learn, and Theano. You can create
visualizations by using Bokeh, Holoviews, Matplotlib, and Datashader while using
Anaconda.

 How to Use Anaconda for Python

Now that we have discussed all the basics in our Python Anaconda
tutorial, let’s discuss some fundamental commands you can use to start using
this package manager. Listing All Environments To begin using Anaconda,
you’d need to see how many Conda environments are present in your
machine.
conda env list - It will list all the available Conda environments in your
machine. Creating a New Environment You can create a new Conda environment
by going to the required directory and use this command:
conda create -n <your_environment_name>

15
You can replace <your_environment_name> with the name of your environment.
After entering this command, conda will ask you if you want to proceed to which
you should reply with y: proceed ([y])/n)?
On the other hand, if you want to create an environment with a particular version
of Python, you should use the following command:
conda create -n <your_environment_name> python=3.6
Similarly, if you want to create an environment with a particular package, you can
use the following command:
conda create -n <your_environment_name>pack_name
Here, you can replace pack_name with the name of the package you want to use.
If you have a .yml file, you can use the following command to create a new Conda
environment based on that file:
conda env create -n <your_environment_name> -f <file_name>.yml
We have also discussed how you can export an existing Conda
environment to a .yml file later in this article.

 Activating an Environment
You can activate a Conda environment by using the following command:
conda activate <environment_name>
You should activate the environment before you start working on the same. Also,
replace the term <environment_name> with the environment name you want to
activate. On the other hand, if you want to deactivate an environment use the
following command:
conda deactivate

 Installing Packages in an Environment

Now that you have an activated environment, you can install packages
into it by using the following command:
conda install <pack_name>
Replace the term <pack_name> with the name of the package you want to install
in your Conda environment while using this command.

16
  Updating Packages in an Environment
If you want to update the packages present in a particular Conda
environment, you should use the following command:
conda update
The above command will update all the packages present in the environment.
However, if you want to update a package to a certain version, you will need to
use the following command:
conda install <package_name>=<version>

 Exporting an Environment Configuration

Suppose you want to share your project with someone else (colleague,
friend, etc.). While you can share the directory on Github, it would have many
Python packages, making the transfer process very challenging. Instead of that,
you can create an environment configuration .yml file and share it with that person.
Now, they can create an environment like your one by using the .yml file.
For exporting the environment to the .yml file, you’ll first have to activate the same
and run the following command:
conda env export ><file_name>.yml
The person you want to share the environment with only has to use the exported
file by using the ‘Creating a New Environment’ command we shared before.

 Removing a Package from an Environment

If you want to uninstall a package from a specific Conda environment, use
the following command:
conda remove -n <env_name><package_name>
On the other hand, if you want to uninstall a package from an activated
environment, you’d have to use the following command:
conda remove <package_name>

 Deleting an Environment
Sometimes, you don’t need to add a new environment but remove one. In
such cases, you must know how to delete a Conda environment, which you can do
so by using the following command:

17
conda env remove –name <env_name>
The above command would delete the Conda environment right away.

4.4 PROJECT MANAGEMENT PLAN

Fig 4.4 – Project management plan

18
 CHAPTER 5

IMPLEMENTATION DETAILS
5.1 ALGORITHMS
We applied the following algorithms to solve this situation:
 natural language processing
 term frequency-inverse document frequency (tf-idf)
 latent dirichlet allocation (lda)

5.1.1 NATURAL LANGUAGE PROCESSING

Natural language processing (NLP) is a field of artificial intelligence in which

computers analyze, understand, and derive meaning from human language in a
smart and useful way. By utilizing NLP, developers can organize and structure
knowledge to perform tasks such as automatic summarization, translation, named
entity recognition, relationship extraction, sentiment analysis, speech recognition,
and topic segmentation. Resume collection is being performed and folder
Structure creation is being done. Data Cleaning such as removing cluster and
unnecessary punctuation would be taken off. Importing necessary libraries is
performed. This library creates a summary of the supplied information within the
word limit. With the help of Natural language processing techniques, the
application which extracts the names of people, places, and other entities from
text, the main goal being to get a reduced version of it that retains the most
important information. It is followed by Exploratory Data Analysis. NLP is used to
analyze text, allowing machines to understand how humans speak. This human-
computer interaction enables real-world applications like automatic text
summarization, sentiment analysis, topic extraction, named entity
recognition, parts-of-speech tagging, relationship extraction, stemming, and more.
NLP is commonly used for text mining, machine translation, and automated
question answering.
A decade ago, the job-seeking and recruiting process were way different
from what it is right now. With the constant and exponential advancement in
technology, the talent acquisition process has changed by leaps and bounds.
Earlier, if an applicant wishes to apply for a job, he needs to drop a resume at the

19
doorstep of the office in the hope of getting called for an interview. But now the
scenario is totally different. The immense increase of internet connectivity has
enabled change in the recruitment process of companies. With the various job
boards and e-recruitment portals, both the recruiters and candidates got some
benefits and convenience but there are a few challenges as well. Earlier, the
companies who received 10–50 resumes on a daily basis, now are flooded with
thousands of resumes. It’s practically impossible and cost-ineffective to go through
these many resumes. Also, there is no standard resume format that the candidate
uses. People with different backgrounds, industries and job profiles have their own
way of presenting details on their resume which is a daunting task for recruiters.
To overcome this, companies and websites started seeking the candidates’ details
through online forms but they also fail to do justice to the process as the forms are
very generic in nature. All these challenges have been substantially achieved by
the Application Tracking System aka ATS which is based on Natural Language
Processing.

Natural Language Processing divides the content mentioned in the resume

into small grammatical parts that is known as parsing. This allows the system to
extract the relevant information from the resume and create a summary of it
irrespective of the order and format of the resume. After creating the summary, the
system ranks the candidates by mapping the similarity between the content
mentioned in their resume and in the job description. So, this is how the NLP is
used in resume screening. Let us walk through the technical depth of the process.
Talking about the Resume Screening using NLP, there are two major phases that
the system works in. The first is Information Extraction and another one is
Candidate Recommendation on the basis of Content. Let us understand these
things.

 INFORMATION EXTRACTION
As we know, there is no standard format for resumes so inconsistencies
and irrelevant data are inevitable. The objective of information extraction is to
extract relevant information or keywords from the unstructured data in the resume
without any human intervention. Using techniques like Tokenisation, Stemming,
POS Tagging etc shown In Fig 4.4 the system can obtain relevant job-related

20
information from the uploaded candidates. The summary of each resume is
created in a JSON format that can be easily used for further processing tasks in
the next phase of this resume screening system.

 TOKENISATION
The process of identifying terms or words that form up a character
sequence is called tokenization. This involved dividing the big text into smaller
parts called tokens; the characters such as whitespace and punctuation are
removed in this process. Through Tokenization, we can get information like the
number of words in a text, frequency of a particular word in the text etc.
Tokenization is an essential step for further text processing such as removal of
stop words, lemmatization and stemming.

 STEMMING AND LEMMATIZATION

In the English language, a single word has multiple forms in multiple
sentences according to its grammatical rules. The task here is to reduce all the
derived or altered forms into their root word. We can overcome this by stemming
and lemmatization but their approaches are different from each other. Stemming is
the mechanism of reducing inflected or derived words to their word root, or stem.
This involves the removal of derivational affixes. On the basis of ruled based
algorithms, the words often undergo over-steaming and under-streaming.
Whereas, lemmatization is the process of using a language dictionary to perform
an accurate reduction to root words. Unlike Stemming that merely cuts off tokens
by easy pattern matching, lemmatization is an additional careful approach that
uses language vocabulary and morphological analysis of words to present
lingually correct lemmas. This implies lemmatization utilises the data of context
and so will completely differentiate between words that have different meanings as
per parts of speech.

 PARTS OF SPEECH (POS) TAGGING

POS is a process in which on the basis of a word’s context in a sentence,
its grammatical information is assigned. The POS tag specifies whether the word
is a noun, pronoun, verb, adjective, etc. as per its usage within the sentence. It’s
crucial to assign these tags to understand the right meaning of a sentence and for

21
building data graphs for named entity recognition. For example: In the sentence “I
am building a website”, here the building is a verb, but in the sentence “I live in the
tallest building of Paris”, here the building is a noun.

5.1.2 TERM FREQUENCY-INVERSE DOCUMENT FREQUENCY (TF-IDF)

It is a numerical statistic that is intended to reflect how important a word is

to a document in a collection or corpus. It is often used as a weighting factor in
searches of information retrieval, text mining, and user modeling. The tf–idf value
increases proportionally to the number of times a word appears in the document
and is offset by the number of documents in the corpus that contain the word,
which helps to adjust for the fact that some words appear more frequently in
general.tf–idf is one of the most popular term-weighting schemes today.

TF-IDF (term frequency-inverse document frequency) is a statistical

measure that evaluates how relevant a word is to a document in a collection of
documents. This is done by multiplying two metrics: how many times a word
appears in a document, and the inverse document frequency of the word across a
set of documents. It has many uses, most importantly in automated text analysis,
and is very useful for scoring words in machine learning. TF-IDF was invented for
document search and information retrieval. It works by increasing proportionally to
the number of times a word appears in a document but is offset by the number of
documents that contain the word. So, words that are common in every document,
such as this, what, and if, rank low even though they may appear many times
since they don’t mean much to that document in particular. However, if the word
Bug appears many times in a document, while not appearing many times in
others, it probably means that it’s very relevant. For example, if what we’re doing
is trying to find out which topics some NPS responses belong to, the word Bug
would probably end up being tied to the topic Reliability, since most responses
containing that word would be about that topic.

TF-IDF for a word in a document is calculated by multiplying two different metrics:

The term frequency of a word in a document. There are several ways of
calculating this frequency, with the simplest being a raw count of instances a word

22
appears in a document. Then, there are ways to adjust the frequency, by the
length of a document, or by the raw frequency of the most frequent word in a
document. The inverse document frequency of the word across a set of
documents. This means, how common or rare a word is in the entire document
set. The closer it is to 0, the more common a word is. This metric can be
calculated by taking the total number of documents, dividing it by the number of
documents that contain a word, and calculating the logarithm. So, if the word is
very common and appears in many documents, this number will approach 0.
Otherwise, it will approach 1. Multiplying these two numbers results in the TF-IDF
score of a word in a document. The higher the score, the more relevant that word
is in that particular document.

Machine learning with natural language is faced with one major hurdle – its
algorithms usually deal with numbers, and natural language is, well, text. So, we
need to transform that text into numbers, otherwise known as text vectorization.
It’s a fundamental step in the process of machine learning for analyzing data, and
different vectorization algorithms will drastically affect end results, so you need to
choose one that will deliver the results you’re hoping for. Once you’ve transformed
words into numbers, in a way that’s machine learning algorithms can understand,
the TF-IDF score can be fed to algorithms such as Naive Bayes and Support
Vector Machines, greatly improving the results of more basic methods like word
counts. This works simply by putting, a word vector represents a document as a
list of numbers, with one for each possible word of the corpus. Vectorizing a
document is taking the text and creating one of these vectors, and the numbers of
the vectors somehow represent the content of the text. TF-IDF enables us to gives
us a way to associate each word in a document with a number that represents
how relevant each word is in that document. Then, documents with similar,
relevant words will have similar vectors, which is what we are looking for in a
machine learning algorithm.

At this stage, a dynamic Script for the Tf-Idf approach is written. Term
frequency-inverse document frequency, is a numerical statistic that is intended to
reflect how important a word is to a document in a collection. The TF-IDF
Victimizer tokenizes texts, learns vocabulary, inverts frequency weightings, and

23
allows encoding new ones. It provides information on a word frequency in the
documents. Higher the TF- IDF score of a term which is computed using the above
equations represents more relevance in a document.

5.1.3 LATENT DIRICHLET ALLOCATION (LDA)

A tool and technique for Topic Modeling, Latent Dirichlet Allocation (LDA)
classifies or categorizes the text into a document and the words per topic, these
are modeled based on the Dirichlet distributions and processes. Latent Dirichlet
Allocation has been used in the application for the following functions:

 Discovering the hidden themes in the data.

 Classifying the data into the discovered themes.
 Using the classification to organize/summarize/search the documents.

The application then deals with the calculation of the score for a candidate’s
resume according to the job posting they have applied for. According to the score
each candidate’s resume receives, a rank list will be made with the candidate
receiving a higher score placed higher as compared to the candidate receiving a
lower score. By displaying a resume list in order of relevance to the position, the
technique ranks CVs according to their match with the job description, making it
easy for recruiters. Customized options for job description in our web application
are shown. Slider options are present in our app for a better user experience. The
web application would also have a Bar Chart that shows the stats. The Resume
Screening System replaces ineffective manual screening, ensuring that no
candidate is overlooked.

5.2 EXISTING SYSTEM

The existing system is a traditional Machine Learning based system. It

gives lower rates of accuracy. It has lower efficiency. It gives lower inaccurate
results. This system may lead to loss of human potential.

24
5.3 PROPOSED SYSTEM

Our system is resume ranking software that uses natural language

processing (NLP) and machine learning. This AI-powered resume screening
program goes beyond keywords to contextually screen resumes. Following
resume screening, the software rates prospects in real-time depending on the
recruiter's job needs. The web application aims to order the resumes, by
intelligently reading job descriptions as input and comparing the resumes which
fall into the category of given Job Descriptions. It provides a ranking after filtering
and recommends the better resume for a given textual job description.

In order to match and rate candidates in real-time, the software employs

natural language processing. Unlike generic processes, this app utilizes Mong for
string matching, Cosine Similarity, Overlapping coefficient Natural Language. Our
work takes a different approach as it focuses mainly on the content of the resumes
where we perform the extraction of skills and related parameters to match
candidates with the job descriptions.The interactive web application will allow the
job applicants to submit their resumes and apply for job postings they may still be
interested in. The resumes submitted by the candidates are then compared with
the job profile requirement posted by the company recruiter by using techniques
like machine learning and Natural Language Processing (NLP).

Scores can then be given to the resumes and they can be ranked from
highest match to lowest match. This ranking is made visible only to the company
recruiter who is interested to select the best candidates from a large pool of
candidates. This is done with the aim to aid the recruiters of any company from the
long and tedious task of viewing and analyzing thousands of candidate resumes.
The calculated ranking scores can then be utilized to determine best-fitting
candidates for that particular job opening Since the dynamic model leverages
NLP, it gives the output instantly. While going through all these pipelines, It will
score each resume and give out accurate output with higher efficiency, precision,
and accuracy. It works the following way.

25
 Resumes from the Data set are parsed to remove white spaces, numbers,
stop words like and, or, etc. TF-IDF vectorization is then applied to convert the
words in the resumes to vectors. The text in the job description is also converted
to vectors using the TF-IDF vectorizer. Cosine distance is computed to measure
the similarity between the resume and the job description provided and Then
Mong String algorithm is applied to identify the resumes which are closely
matching with the JD provided by the recruiters.The main contributions of our work
can be summarized as follows: Use of our framework is not limited to a single field
of application and is useful for many more real-world applications. Providing a
ranking-based approach after filtering Framework to highlight skills of a resume

26
 CHAPTER 6

RESULTS AND DISCUSSION

In this document, we developed an automated resume screening system

that simplifies the e-recruitment process by reducing the numerous issues that
recruiters faced when they relied on human shortlisting of candidates for a given
job posting. The model was generated with high accuracy using the Kneighbors
approach.

Fig.6.1 Kneighbors classified is constructed

Using Kneighbors algorithm, we have got 98% of accuracy which is the perfect
accuracy which is shown in Fig 6.1. On first glance, Linear SVC and XGBoost
provided the highest accuracy and a favorable confusion matrix. Both were able to
correctly forecast the Role for a specific job description given to the system.

27
 Fig.6.2 confusion matrix is plotted

The issue arose during calculating the probability distribution of the

projected findings and the confusion matrix is plotted and shown in Fig 6.2. These
two algorithms assigned a probability to all roles, no matter how small the number,
which caused a difficulty when evaluating our resumes using the method
described in Resume Filtering and Resume Screening. The KNN method provided
poor accuracy, but the probability distribution was appropriate for our needs.

Fig 6.3 Word Cloud Representing Most Common occurring words

The top 50 most frequently used words chosen described each job
description efficiently. These words helped train our model and provide
satisfactory results. We plotted Word Cloud to visualize the most common
occurring words presented in fig. 7.3

28
 CHAPTER 7

SUMMARY

7.1 CONCLUSION
Our algorithm was successfully able to screen and shortlist the best
candidates with the help of NLP.Highly accurate results were obtained by using
Latent Dirichlet Allocation to display the best-shortlisted resume on Web Ui. The
web application was successfully able to order the resumes, by intelligently
reading job descriptions as input and comparing the resumes which fall into the
category of given Job Descriptions. The organization receives a large number of
applications for each employment opening. Finding the right candidate's
application from a sea of resumes is a time-consuming endeavor for any company
these days. The classification of a candidate's resume is a laborious, time-
consuming, and resource-intensive process. To address this problem, we created
an automated machine learning-based algorithm that recommends acceptable
applicant resumes to HR based on the job description provided. The suggested
methodology had two stages: first, it classified resumes into various groups.
Second, it suggests resumes based on their resemblance to the job description. If
an industry produces a high number of resumes, the proposed approach can be
used to create an Industry-specific model.

By engaging domain experts such as HR professionals, a more accurate

model may be built, and HR professionals' feedback can be used to iteratively
enhance the model. The results from the model are encouraging. The resume
classifier application is successful in automating the manual task of project
allocation to the new recruits of the organization based on the interests, work
experience, and expertise mentioned by the candidate in the profile. The Resume
Screening System replaces ineffective manual screening, ensuring that no
candidate is overlooked. The need for efficient and effective resume screening is
at the heart of every excellent recruitment strategy. The system will be able to
accept or reject a job applicant based on two factors: the company's requirements
must match the skills listed in the applicant's resume, and the test evaluation will

29
be based on the applicant's skills, ensuring that the resumes uploaded by the
applicant are genuine and the applicant is truly
knowledgeable about the skills. Using NLP(Natural Language Processing) and
ML(Machine Learning) to rank the resumes according to the given constraint, this
intelligent system ranks the resume of any format according to the given
constraints or the following requirement provided by the client company. We will
basically take the bulk of input resume from the client company and that client
company will also provide the requirement and the constraints according to which
the resume should be ranked by our system. Besides the information provided by
the resume, we are going to read the candidate’s social profiles (like LinkedIn,
GitHub, etc.) which will give us more genuine information about that candidate.
The application automates the task of project allocation, thereby eliminating the
tedious and redundant affair of opening and analyzing the resumes manually by
the HR team of the organization.

7.2 FUTURE WORK

The application can be extended further to other domains like Telecom,
Healthcare, E-commerce, and public sector jobs. We also wish to put into effect
and present a smart evaluation in the consistent database to survey with the
present models. Efforts can be made to explore whether it is possible to identify in
advance what lists might be ranked worse than the current baseline and to
investigate whether there is another way to transform word embedding’s into
document embedding’s, such as using doc2vec, that could have a greater impact
on LTR results. Another area we would want to focus on is ranking the resumes of
people in different languages. Be it Dutch, German, French, or Hindi, we want to
explore further ways to analyze and operate our model on major languages of the
world.

30
 REFERENCES

[1] Abeer Zaroor; Mohammed Maree; Muath Sabha,” A Hybrid Approach to

Conceptual Classification and Ranking of Resumes and Their Corresponding Job
Post”, Smart Innovation, Systems and Technologies book series (SIST, volume
72), 2017

[2] Anushka Sharma; Smiti Singhal; Dhara Ajudia,” Intelligent Recruitment System
Using NLP”, International Conference on Artificial Intelligence and Machine Vision
(AIMV), 2021

[3] Fahad, SK Ahmed, and Abdul Samad Ebrahim Yahya. "Inflectional review of
deep learning on natural language processing." 2018 International Conference on
Smart Computing and Electronic Enterprise (ICSEE). IEEE, 2018

[4] Garima Bhardwaj; S. Vikram Singh; Vinay Kumar,” An Empirical Study of

Artificial Intelligence and its Impact on Human Resource Functions”, International
Conference on Computation, Automation and Knowledge Management (ICCAKM),
2020

[5] Jitendra Purohit; Aditya Bagwe; Rishabh Mehta; Ojaswini Mangaonkar;

Elizabeth George, “Natural Language Processing based Jaro-The Interviewing
Chatbot”, 2019 3rd International Conference on Computing Methodologies and
Communication (ICCMC), August 2019

[6] Mujtaba, Dena F., and Nihar R. Mahapatra. "Ethical Considerations in AI-Based
Recruitment." 2019 IEEE International Symposium on Technology and Society
(ISTAS). IEEE, 2019.

[7] Pradeep Kumar Roy; Sarabjeet Singh Chowdhary; Rocky Bhatia,” A Machine
Learning approach for automation of Resume Recommendation system”, Procedia
Computer Science Volume 167, Pages 2318-2327, 2020

31
[8] Rajath V; Riza Tanaz Fareed; Sharadadevi Kaganurmath,” Resume
Classification And Ranking Using KNN And Cosine Similarity”,international
JOURNAL OF ENGINEERING RESEARCH & TECHNOLOGY, Volume 10, Issue
08, AUGUST 2021

[9] Tejaswini K; Umadevi V; Shashank M Kadiwal; Sanjay Revanna,” Design and

Development of Machine Learning based Resume Ranking System”, Global
Transitions Proceedings, October 2021

[10] Yong Luo; Huaizheng Zhang; Yongjie Wang; Yonggang Wen; Xinwen Zhang,”
ResumeNet: A Learning-Based Framework for Automatic Resume Quality
Assessment”, 2018 IEEE International Conference on Data Mining (ICDM),
December 2018

32
 APPENDIX
A. SOURCE CODE

import matplotlib.colors as mcolors

import gensim
import gensim.corpora as corpora
from operator import index
from wordcloud import WordCloud
from pandas._config.config import options
import pandas as pd
import streamlit as st
import plotly.express as px
import plotly.graph_objects as go
import matplotlib.pyplot as plt
import Similar
from PIL import Image
import time

image = Image.open('Images//logo.png')
st.image(image, use_column_width=True)
st.title("Resume Screening")
# Reading the CSV files prepared by the fileReader.py
Resumes = pd.read_csv('Resume_Data.csv')
Jobs = pd.read_csv('Job_Data.csv')
#############JOB DESCRIPTION CODE ##############
# Checking for Multiple Job Descriptions
# If more than one Job Descriptions are available, it asks user to select one as
well.
if len(Jobs['Name']) <= 1:
st.write(
"There is only 1 Job Description present. It will be used to create scores.")
else:
st.write("There are ", len(Jobs['Name']),

33
 "Job Descriptions available. Please select one.")

# Asking to Print the Job Desciption Names

if len(Jobs['Name']) > 1:
option_yn = st.selectbox(
"Should we need to Show the Job Description Names?", options=['YES',
'NO'])
if option_yn == 'YES':
index = [a for a in range(len(Jobs['Name']))]
fig = go.Figure(data=[go.Table(header=dict(values=["Job No.", "Job Desc.
Name"], line_color='darkslategray',
fill_color='lightskyblue'),
cells=dict(values=[index, Jobs['Name']],
line_color='darkslategray',
fill_color='cyan'))
])
fig.update_layout(width=700, height=400)
st.write(fig)
# Asking to chose the Job Description
index = st.slider("Slide to select te JD : ", 0,
len(Jobs['Name'])-1, 1)
option_yn = st.selectbox("Do you want Job Description to be displayed ?",
options=['YES', 'NO'])
if option_yn == 'YES':
st.markdown("---")
st.markdown("### Job Description :")
fig = go.Figure(data=[go.Table(
header=dict(values=["Job Description"],
fill_color='#f0a500',
align='center', font=dict(color='white', size=16)),
cells=dict(values=[Jobs['Context'][index]],
fill_color='#f4f4f4',

34
 align='left'))])
fig.update_layout(width=800, height=500)
st.write(fig)
st.markdown("---")

############### SCORE CALCUATION ####################

@st.cache()
def calculate_scores(resumes, job_description):
scores = []
for x in range(resumes.shape[0]):
score = Similar.match(
resumes['TF_Based'][x], job_description['TF_Based'][index])
scores.append(score)
return scores

Resumes['Scores'] = calculate_scores(Resumes, Jobs)

Ranked_resumes = Resumes.sort_values(
by=['Scores'], ascending=False).reset_index(drop=True)

Ranked_resumes['Rank'] = pd.DataFrame(
[i for i in range(1, len(Ranked_resumes['Scores'])+1)])

#################### SCORE TABLE PLOT ##################

fig1 = go.Figure(data=[go.Table(
header=dict(values=["Rank", "Name", "Scores"],
fill_color='#00416d',
align='center', font=dict(color='white', size=16)),
cells=dict(values=[Ranked_resumes.Rank, Ranked_resumes.Name,
Ranked_resumes.Scores],
fill_color='#d6e0f0',
align='left'))])

fig1.update_layout(title="Top Ranked Resumes", width=700, height=1100)

35
st.write(fig1)

st.markdown("---")

fig2 = px.bar(Ranked_resumes,
x=Ranked_resumes['Name'], y=Ranked_resumes['Scores'],
color='Scores',
color_continuous_scale='haline', title="Score and Rank Distribution")
# fig.update_layout(width=700, height=700)
st.write(fig2)
st.markdown("---")

#################### TF-IDF Code ####################

@st.cache()
def get_list_of_words(document):
Document = []

for a in document:
raw = a.split(" ")
Document.append(raw)

return Document
document = get_list_of_words(Resumes['Cleaned'])

id2word = corpora.Dictionary(document)
corpus = [id2word.doc2bow(text) for text in document]

lda_model = gensim.models.ldamodel.LdaModel(corpus=corpus,
id2word=id2word, num_topics=6, random_state=100,
update_every=3, chunksize=100, passes=50,
alpha='auto', per_word_topics=True)

################## LDA CODE #######################

@st.cache # Trying to improve performance by reducing the rerun computations

36
def format_topics_sentences(ldamodel, corpus):
sent_topics_df = []
for i, row_list in enumerate(ldamodel[corpus]):
row = row_list[0] if ldamodel.per_word_topics else row_list
row = sorted(row, key=lambda x: (x[1]), reverse=True)
for j, (topic_num, prop_topic) in enumerate(row):
if j == 0:
wp = ldamodel.show_topic(topic_num)
topic_keywords = ", ".join([word for word, prop in wp])
sent_topics_df.append(
[i, int(topic_num), round(prop_topic, 4)*100, topic_keywords])
else:
break

return sent_topics_df

################ Topic Word Cloud Code #####################

# st.sidebar.button('Hit Me')
st.markdown("## Topics and Topic Related Keywords ")
st.markdown(
"""This Wordcloud representation shows the Topic Number and the Top
Keywords that contstitute a Topic.
This further is used to cluster the resumes. """)

cols = [color for name, color in mcolors.TABLEAU_COLORS.items()]

cloud = WordCloud(background_color='white',
width=2500,
height=1800,
max_words=10,
colormap='tab10',
collocations=False,
color_func=lambda *args, **kwargs: cols[i],

37
 prefer_horizontal=1.0)

topics = lda_model.show_topics(formatted=False)

fig, axes = plt.subplots(2, 3, figsize=(10, 10), sharex=True, sharey=True)

for i, ax in enumerate(axes.flatten()):
fig.add_subplot(ax)
topic_words = dict(topics[i][1])
cloud.generate_from_frequencies(topic_words, max_font_size=300)
plt.gca().imshow(cloud)
plt.gca().set_title('Topic ' + str(i), fontdict=dict(size=16))
plt.gca().axis('off')

plt.subplots_adjust(wspace=0, hspace=0)
plt.axis('off')
plt.margins(x=0, y=0)
plt.tight_layout()
st.pyplot(plt)

st.markdown("---")

### SETTING UP THE DATAFRAME FOR SUNBURST-GRAPH #############

df_topic_sents_keywords = format_topics_sentences(
ldamodel=lda_model, corpus=corpus)
df_some = pd.DataFrame(df_topic_sents_keywords, columns=[
'Document No', 'Dominant Topic', 'Topic % Contribution',
'Keywords'])
df_some['Names'] = Resumes['Name']

df = df_some

38
st.markdown("## Topic Modeling with Resume using LDA ")
st.markdown(
"Using LDA to divide the topics into a number of usefull topics and creating a
Cluster of matching topic resumes. ")
fig3 = px.sunburst(df, path=['Dominant Topic', 'Names'], values='Topic %
Contribution',
color='Dominant Topic', color_continuous_scale='viridis', width=800,
height=800, title="Topic Distribution Graph")
st.write(fig3)

############ RESUME PRINTING

option_2 = st.selectbox("Do you want to see the Best Resumes?", options=[
'YES', 'NO'])
if option_2 == 'YES':
indx = st.slider("Slide, Which resume to display ?:",
1, Ranked_resumes.shape[0], 1)

st.write("Displaying Resume with Rank: ", indx)

st.markdown("---")
st.markdown("## **Resume** ")
value = Ranked_resumes.iloc[indx-1, 2]
st.markdown("#### The Word Cloud For Resume")
wordcloud = WordCloud(width=800, height=800,
background_color='white',
colormap='viridis', collocations=False,
min_font_size=10).generate(value)
plt.figure(figsize=(7, 7), facecolor=None)
plt.imshow(wordcloud)
plt.axis("off")
plt.tight_layout(pad=0)
st.pyplot(plt)

39
 st.write("With a Match Score of :", Ranked_resumes.iloc[indx-1, 6])
fig = go.Figure(data=[go.Table(
header=dict(values=["Resume"],
fill_color='#f0a500',
align='center', font=dict(color='white', size=16)),
cells=dict(values=[str(value)],
fill_color='#f4f4f4',
align='left'))])

fig.update_layout(width=800, height=1200)
st.write(fig)
st.markdown("---")
import spacy
import Distill
try:
nlp = spacy.load('en_core_web_sm')

except ImportError:
print("Spacy's English Language Modules aren't present \n Install them by doing
\n python -m spacy download en_core_web_sm")

def _base_clean(text):
"""
Takes in text read by the parser file and then does the text cleaning.
"""
text = Distill.tokenize(text)
text = Distill.remove_stopwords(text)
text = Distill.remove_tags(text)
text = Distill.lemmatize(text)
return text

def _reduce_redundancy(text):
"""

40
 Takes in text that has been cleaned by the _base_clean and uses set to reduce
the repeating words
giving only a single word that is needed.
"""
return list(set(text))

def _get_target_words(text):
"""
Takes in text and uses Spacy Tags on it, to extract the relevant Noun, Proper
Noun words that contain words related to tech and JD.
"""
target = []
sent = " ".join(text)
doc = nlp(sent)
for token in doc:
if token.tag_ in ['NN', 'NNP']:
target.append(token.text)
return target

def Cleaner(text):
sentence = []
sentence_cleaned = _base_clean(text)
sentence.append(sentence_cleaned)
sentence_reduced = _reduce_redundancy(sentence_cleaned)
sentence.append(sentence_reduced)
sentence_targetted = _get_target_words(sentence_reduced)
sentence.append(sentence_targetted)
return sentence
import nltk
import spacy
import re

from nltk.tokenize import word_tokenize, sent_tokenize

from nltk.corpus import stopwords

41
# Define english stopwords
stop_words = stopwords.words('english')
# load the spacy module and create a nlp object
# This need the spacy en module to be present on the system.
nlp = spacy.load('en_core_web_sm')
# proces to remove stopwords form a file, takes an optional_word list
# for the words that are not present in the stop words but the user wants them
deleted.

def remove_stopwords(text, stopwords=stop_words, optional_params=False,

optional_words=[]):
if optional_params:
stopwords.append([a for a in optional_words])
return [word for word in text if word not in stopwords]

def tokenize(text):
# Removes any useless punctuations from the text
text = re.sub(r'[^\w\s]', '', text)
return word_tokenize(text)

def lemmatize(text):
# the input to this function is a list
str_text = nlp(" ".join(text))
lemmatized_text = []
for word in str_text:
lemmatized_text.append(word.lemma_)
return lemmatized_text

# internal fuction, useless right now.

def _to_string(List):
# the input parameter must be a list
string = " "

42
 return string.join(List)

def remove_tags(text, postags=['PROPN', 'NOUN', 'ADJ', 'VERB', 'ADV']):

"""
Takes in Tags which are allowed by the user and then elimnates the rest of the
words
based on their Part of Speech (POS) Tags.
"""
filtered = []
str_text = nlp(" ".join(text))
for token in str_text:
if token.pos_ in postags:
filtered.append(token.text)
return filtered
from operator import index
from pandas._config.config import options
import Cleaner
import textract as tx
import pandas as pd
import os
import tf_idf

resume_dir = "Data/Resumes/"
job_desc_dir = "Data/JobDesc/"
resume_names = os.listdir(resume_dir)
job_description_names = os.listdir(job_desc_dir)

document = []

def read_resumes(list_of_resumes, resume_directory):

placeholder = []
for res in list_of_resumes:
temp = []
temp.append(res)

43
 text = tx.process(resume_directory+res, encoding='ascii')
text = str(text, 'utf-8')
temp.append(text)
placeholder.append(temp)
return placeholder

document = read_resumes(resume_names, resume_dir)

def get_cleaned_words(document):
for i in range(len(document)):
raw = Cleaner.Cleaner(document[i][1])
document[i].append(" ".join(raw[0]))
document[i].append(" ".join(raw[1]))
document[i].append(" ".join(raw[2]))
sentence = tf_idf.do_tfidf(document[i][3].split(" "))
document[i].append(sentence)
return document

Doc = get_cleaned_words(document)
Database = pd.DataFrame(document, columns=[
"Name", "Context", "Cleaned", "Selective", "Selective_Reduced",
"TF_Based"])
Database.to_csv("Resume_Data.csv", index=False)

def read_jobdescriptions(job_description_names, job_desc_dir):

placeholder = []
for tes in job_description_names:
temp = []
temp.append(tes)
text = tx.process(job_desc_dir+tes, encoding='ascii')
text = str(text, 'utf-8')
temp.append(text)
placeholder.append(temp)
return placeholder

44
job_document = read_jobdescriptions(job_description_names, job_desc_dir)

Jd = get_cleaned_words(job_document)
jd_database = pd.DataFrame(Jd, columns=[
"Name", "Context", "Cleaned", "Selective", "Selective_Reduced",
"TF_Based"])
jd_database.to_csv("Job_Data.csv", index=False)

45
 B. SCREENSHOTS

Score and rand distribution

Data visualization of word that were used more repeatedly in the

resume collected

46
 We have build and got the perfect accuracy

We have build an web application code with the algorithm which has
given best performance.It has an accuracy of 98.3%.

47
 Pie chart is plotted

Plotting the confusion matrix

48