E-Recruiting and shortlisting using Candidate Resume

with NLP and Machine Learning

Sowndharya R1,a), Sneha T2,b)
1
Assistant Professor, Department of CSE, Sona College of Technology ,Salem, India.
2
UG Scholar, Department of CSE, Sona College of Technology, Salem, India.
a)
[email protected]
b)
[email protected]

Abstract:

The project aims to solve the problems of today's business world by publishing an iterative smart sorting system involving
experts, designed specifically for use in recruiting people. Unlike existing software, this solution prioritizes data privacy,
improves security measures and makes the system more robust. The system is optimized for job recruiters and the candidate
evaluation process, allowing recruiters to identify suitable candidates. The system increases the efficiency of the
recruitment process by semantically analyzing resumes and leveraging predefined ontologies. This research leads to a new
combination of expertise and technology to deliver effective solutions to recruiters based on the complexity of candidates
competing in a dynamic marketplace.

Keywords: Machine learning, natural language processing (NLP), pattern recognition, classification, supervised learning,
Artificial intelligence.

1. INTRODUCTION
This paper addresses the imperative need to streamline the time-intensive process of analyzing numerous
resumes for a single job role, a task often undertaken by human resources professionals. Traditional methods
on job posting platforms or job boards necessitate extensive hours of candidate assessment and recruitment.
The proposed system is designed to significantly reduce the time spent by HR professionals, providing a more
efficient means of evaluating resumes and expediting the recruitment process. This paper elucidates how the
implementation of this system aligns with industry requirements, offering a prompt and effective solution to
the challenges posed by the overwhelming volume of resumes in the assessment and recruitment landscape.

Resume Tracking Module is a solution designed to improve the hiring process through multiple advertising
objectives. First, it aims to make it easier for candidates to continue using a storage and recovery tool.
Automation is the focus and the model incorporates technology to extract data generated through repetitive
work, reducing manual work. The system uses optical character recognition (OCR) to perform objective skills
analysis and assign scores to assess candidate suitability. Ensuring that employees are successful and
improving the accuracy of shortlisting candidates according to the manager's needs is also an important goal.
Capable of processing up to 100 resumes simultaneously, the module simplifies the job and speeds up the first
stage of recruitment. By reducing the number of manual controls, working time is further shortened and
decisions are made faster. A user-friendly interface is key to ease of navigation for recruiters, and integration
into existing jobs ensures seamless integration of the recruitment process. Finally, this model provides accurate
and timely information, providing recruiters and recruiters with valuable input to make informed decisions
throughout the hiring process. It essentially combines technology development with a user-oriented approach
to create efficient, accurate and user-friendly products.

2. PROPOSED FRAMEWORK
The current recruitment and selection process in our Human Resources Management cycle for technical
projects are plagued by inefficiencies, primarily stemming from the resource-intensive nature of manually
analyzing candidate resumes. The need for substantial manpower and extensive time investments to accurately
shortlist 20 candidates poses a significant challenge, resulting in delays in the hiring process and potential
oversight in identifying the most qualified individuals. This resource-intensive
approach not only strains operational capacity but also hinders the timely acquisition of skilled talent vital for
the success of our technical projects.
The existing recruitment system relies on a manual form-filling process, consuming significant time for
candidates, yet often resulting in job placements that do not align with their skills and preferences. The current
ratio of 5:1, indicating that only one out of five individuals is content with their job, highlights a critical
disconnect between candidate skills and job assignments.For instance, a proficient Python developer may find
themselves working in Java,rendering their specialized skills underutilized. This incongruence between
candidate expertise and job roles leads to dissatisfaction and suboptimal productivity. There is an urgent need
for an innovative HR solution that acts as a bridge between candidates seeking roles aligned with their skills
and companies aiming to secure the best-fit candidates for optimal performance and job satisfaction.

3. IMPLEMENTATION AND WORK-FLOW

Our main goal in language processing (NLP) is to analyze the difference between human language through the
combination of machine learning techniques, management process and accounting process. The program
includes basic NLP functions such as tokenization, lemmatization, stemming, parsing, part of speech tagging,
and language analysis, which work together to break down small words that are the important product.
According to the study model, detailed and preliminary data collection was first carried out in the project in
order to facilitate the identification and analysis of the data. Text is represented using word embeddings and
feature engineering is used to improve model understanding. The important decision in choosing the
appropriate model according to the nature of the job is time. The final step involves rigorous testing of the
selected model and its deployment into a real application. Throughout the process, documents detail each step,
from preliminary sketches to agreement and model determination, promoting a better understanding of the
business.

Ensuring successful extraction from PDF and DOCX files by carefully performing the extraction using
powerful tools such as pdfminer.six and docx2txt. For DOC files, the integration of catdoc as a command line
tool enables complete deletion. Names are extracted using advanced language processing (NLP), specifically
using the nltk library to identify names. The main goal here is to identify areas related to people (PERSON
areas) and consider continuous improvement. To extract phone numbers accurately, a good traditional pattern
matching method is used to identify and extract phone numbers in the reverse code. Similar to phone number
extraction, email address extraction relies on regular reporting standards to ensure that the process and system
for verifying and removing reprints is accurate. When removing intelligence, the operation uses nltk to tokenize
and remove the interception message, along with careful retention of the intelligence database for detailed
information, thus increasing the accuracy of skills extraction. In academic research, the so-called recognition
work of nltk is used only for organizations related to organizations (ORGANIZATION entities). Additionally,
a method can also be used to identify specific words related to institutions, such as "school" or "college," which
are useful in displaying academic article information. Challenges and limitations are acknowledged while
verifying the effectiveness of the study, and external solutions are offered to improve the accuracy of extraction
techniques.

Creating a feedback loop using language processing (NLP) in Python. First, it checks the file extension (.pdf
or docx ,doc ) and converts the content to text. The parser uses NLP libraries like spacy and NLTK to extract
basic information like name, phone number, email ID, qualifications, and skills from resumes.NLP libraries
used:

3.1 Text Capture:

An Optical Character Recognition (OCR) module is included in addition to software capture technology to
clarify the path. OCR helps convert reverse copies of images or text into machine-encoded text. This extends
the integration process for applicants to submit hard copies. The OCR process enables integration with natural
language processing (NLP) by converting digital formats such as PDF and JPG into text.

3.2 Natural Language Processing (NLP):

NLP plays an important role in data recovery. This process includes:
 • Lexical analysis: Break white text into words and phrases, create tags.
• Syntax Analysis: Analyze the grammar and content of words.
• Semantic analysis: Determine the literal meaning of a word.
• Named Entity Recognition (NER): Addresses specific context by identifying real-world objects in the
context.

NER fine-tunes the NLP module to distinguish terms based on context, thus improving the recognition and
interpretation of certain words.

3.3 Candidate ranking and Dashboard Interface:

The web portal allows HR to define the required parameters and skills and embed them into ElasticSearch to
create dashboards. This dashboard has rich charts and graphs integrated to help the HR department plan queries
according to the company's needs. LogStash supports many file types. ElasticSearch queries and ranks resumes
based on constraints set by HR, using the most important factors to score. The final list will be sent to HR for
successful candidate evaluation ..

4. VISUAL DOCUMENTATION

Flow Diagram

FIGURE 1

Frontend

• Login with credential

• Enter the details for the requirement on the specific company ,That looking for number of candidates.
• Uploading the requirement with a given range.
• The required skillset will be parsed i.e, Skills, Scores, Achievements ,etc.,

Resume Parsing & Shortlisting

• Upload the resumes
• Parse the resumes with the required data structured with it using OCR and store it JSON with specific ID.
• If the required set is found in the resume ,Then it’ll be shortlisted.
• Else, It will be moved to draft for secondary verification.
• On the secondary verification, Only the mandatory field will be verified.
• If found It will be shortlisted.
• After secondary verification ,If the needed fields since not found ,Then move the resume to draft for not
shortlisted.

Entity Relationship Model(ER Diagram)

FIGURE 2

Shortlisting Flow-Chart

FIGURE 3

5.CONCLUSION
In conclusion, this research paper underscores the critical need for efficiency in the labor-intensive process of
resume analysis within the realm of HR professionals. By addressing the challenges associated with traditional
methods on job posting platforms, the proposed Automated Resume Analysis System emerges as a timely
solution. The outlined advantages, including enhanced accuracy, streamlined processing, and effective skillset
analysis, position the system as a valuable asset for HR professionals. The objectives of the Resume Tracking
Module, focusing on efficient data handling, automated parsing, skillset analysis, alignment with hiring needs,
and bulk processing, collectively contribute to the system's effectiveness. The emphasis on time-efficient
analysis further highlights the system's commitment to expediting the recruitment process. In essence, this
research provides a comprehensive understanding of the proposed system's capabilities, aligning with industry
needs and offering a promising avenue for revolutionizing resume assessment and recruitment practices.

6. REFERENCE
[1] Akilandeswari, J. and Jothi, G, Performance Comparison of Machine Learning Algorithms that Predicts
Students’ Employability (November 15, 2017). Proceedings of the International Conference on Intelligent
Computing Systems (ICICS 2017 – Dec 15th - 16th 2017) organized by Sona College of Technology, Salem,
Tamilnadu, India, Available at SSRN: https://ssrn.com/abstract=3134357 or
http://dx.doi.org/10.2139/ssrn.3134357

[2]Suresh, Y. and Roshni, S and Senthilkumar, J and V, Mohanraj, Survey on Algorithms for Energy Efficient
Cluster Head Selection in Wireless Sensor Networks (November 15, 2017). Proceedings of the International
Conference on Intelligent Computing Systems (ICICS 2017 – Dec 15th - 16th 2017) organized by Sona
College of Technology, Salem, Tamilnadu, India, Available at
SSRN: https://ssrn.com/abstract=3134234 or http://dx.doi.org/10.2139/ssrn.3134234

[3] Yuvarajan, V. and Sathiyabhama, B and Udhaya Kumar, S, A Comparison of Machine Learning
Techniques for Survival Prediction in Breast Cancer Gene Expression Data (November 15, 2017). Proceedings
of the International Conference on Intelligent Computing Systems (ICICS 2017 – Dec 15th - 16th 2017)
organized by Sona College of Technology, Salem, Tamilnadu, India, Available at
SSRN: https://ssrn.com/abstract=3126112 or http://dx.doi.org/10.2139/ssrn.3126112

[4] Sangeetha, C and Sathyamoorthi, V, Decision Support System for Agricultural Crop Prediction Using
Machine Learning Techniques (November 15, 2017). Proceedings of the International Conference on
Intelligent Computing Systems (ICICS 2017 – Dec 15th - 16th 2017) organized by Sona College of
Technology, Salem, Tamilnadu, India, Available at SSRN: https://ssrn.com/abstract=3134288 or
http://dx.doi.org/10.2139/ssrn.3134288

[5 ]Krishnaveni, V and Anitha Elavarasi, S, Prediction of Disease Severity Using Machine Learning
Algorithm (November 15, 2017). Proceedings of the International Conference on Intelligent Computing
Systems (ICICS 2017 – Dec 15th - 16th 2017) organized by Sona College of Technology, Salem, Tamilnadu,
India, Available at SSRN: https://ssrn.com/abstract=3134279 or http://dx.doi.org/10.2139/ssrn.3134279

[6] Valdez-Almada et al. “Natural Language Processing and Text Mining to Identify Knowledge Profiles for
Software Engineering Positions: Generating Knowledge Profiles from Resumes.” 2017 5th International
Conference in Software Engineering Research and Innovation (CONISOFT) (2017): 97-106. (2017).