Proceedings of the Second International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI 2024)

IEEE Xplore Part Number : CFP24VG0-ART ; ISBN : 979-8-3315-4066-1

Resspar: AI-Driven Resume Parsing and

Recruitment System using NLP and Generative AI
2024 Second International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI) | 979-8-3315-4066-1/24/$31.00 ©2024 IEEE | DOI: 10.1109/ICOICI62503.2024.10696451

Abisha D Keerthana S Navedha Evanjalin R

Department of Computer Science and Department of Computer Science and Department of Computer Science and
Engineering Engineering Engineering
National Engineering College National Engineering College National Engineering College
Kovilpatti, India Kovilpatti, India Kovilpatti, India
[email protected] [email protected] [email protected]

Kavitha K Jothi Mary S Ramya R

Department of Computer Science and Department of Computer Science and Department of Computer Science and
Engineering Engineering Engineering
National Engineering College National Engineering College National Engineering College
Kovilpatti, India Kovilpatti, India Kovilpatti, India
[email protected] [email protected] [email protected]

Abstract—Artificial Intelligence (AI) is a highly emerging for companies across industries. However, the traditional
domain in the current scenario. It has numerous applications in manual review of resumes can be time-consuming, error-
various fields and almost every domain started integrating with prone, and subject to biases. To address these challenges,
AI for better and efficient working. In today’s scenario resumes Resspar System is designed to automate the extraction and
play a vital role as it decides the candidate’s future scope. The analysis of relevant information from resumes, significantly
candidate’s resume must clearly contain the skills and details reducing the time spent manually sifting through resumes,
expected by the recruiters. Resume parser is always a enabling recruiters to focus on assessing the most relevant
demanding field for both the applicants and the recruiters. The applicants. Resspar is a web-based application designed to
LLM model takes in prompts or instructions and generates text
parse and store data from resumes to filter the eligible
that corresponds to the relevant information extracted from the
candidates for the required job. It is an AI and full stack
resume images. This text generation capability is crucial for
parsing and understanding the content of resumes in a
integrated website to meet with the recruiters’ demand on the
structured manner. Generative AI is utilized through the GenAI resume parsing needs. Resspar has a login and register page
API provided by Google. The GenAI API is used for tasks such in which the users can login using their email and password,
as generating text from images of resumes, where it interprets if already have an account. The user login data is stored to the
the visual information and produces structured text output SQLite database and updates the database by inserting the
containing relevant details like names, emails, phone numbers, values as a new login is registered. The home of the resspar
and skills. Resspar aims to streamline the hiring process by can be accessed after logging in. It contains the basic
developing a web-based Resume Parsing System using Natural information about Resspar and its Features and the Upload,
Language Processing (NLP) like Language Model (LLM), then Resources, About, Contact tabs on the right top corner of the
it also uses GenAI, and Prompt Engineering Techniques with page. A short description about resume parsing is given in the
Python and Flask as the backend Framework. It provides a home page and the buttons for the upload page is given in the
user-friendly platform that automates the extraction of essential main section. Below that the Resources, About and contact
information such as personal information (name, email, phone details are provided. The Upload page allows the user to
number) and professional skills from uploaded resumes. select and upload a resume in PDF format. The parsed data
Recruitment processes often involve sifting through a large points are then presented in a structured tabular format,
number of resumes to identify suitable candidates for specific making it easy for recruiters and hiring managers to review
job roles or domains. Resspar’s functionality includes a user and compare candidate profiles. These parsed data are stored
interface for uploading resumes, parsing them using advanced in the SQLite database. Additionally, Resspar includes a
algorithms, storing parsed data in a SQLite database, and
powerful filtering feature. Users can specify job roles or
offering a filtering mechanism to match candidates with specific
job requirements. It identifies the candidates whose skills align
domains, and the system identifies candidates whose skills
with the given criteria. This functionality significantly reduces align with the given criteria. By integrating these features, the
the time spent manually sifting through resumes, enabling goal is to create a seamless and efficient solution for
recruiters to focus on assessing the most relevant applicants. recruiters and HR professionals to identify qualified
Keywords—Resume parser, Natural language Processing, Large candidates quickly and accurately. The filter feature provides
Language Model, Prompt Engineering, GenAI, Python, Flask, an input section in which the user can specifically search a
SQLite, Information extraction skill or job roles to match candidates with specific job
requirements. This can be used by the Recruiting teams to
I. INTRODUCTION find the fittest candidate. Resspar promotes fairness in the
In today’s competitive job market, efficient recruitment hiring process by objectively evaluating candidates based on
processes are crucial for organizations seeking the right their qualifications. It is believed that this innovative solution
talent. Recruitment is a critical aspect of organizational will revolutionize recruitment practices and contribute to
growth, and identifying the right talent remains a top priority building stronger, more diverse teams.

979-8-3315-4066-1/24/$31.00 ©2024 IEEE 1584

Authorized licensed use limited to: Gannon University. Downloaded on December 13,2024 at 04:10:19 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Second International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI 2024)
IEEE Xplore Part Number : CFP24VG0-ART ; ISBN : 979-8-3315-4066-1

II. LITERATURE SURVEY conducted a systematic review of text-stemming methods

A framework was presented by S. Ashrafi et al. (2023) [1] from 1968 to 2023, addressing the growing need for advanced
that makes use of automatic resume analysis to suggest stemming systems due to the exponential growth of
acceptable new jobs with higher pay and the most effective unstructured digital data. The paper categorizes stemming
fast re-education path leading to those jobs. Custom-built algorithms into linguistic knowledge-based, statistical,
deep neural networks based on CNN-Random along with corpus-based, context-sensitive, and hybrid types, analyzing
customized natural language processing tools are designed their methodologies, data sources, performance, and
for large-scale automatic recommendation of a personalized evaluation methods. The study emphasizes the effectiveness
education and career path to each job seeker. The proposed of linguistic knowledge-based stemmers for highly inflected
work is focused on software engineering job search and languages and discusses the challenges and research gaps in
resumes upgrades. Additionally, a book recommendation existing text-stemming approaches. Additionally, the paper
module is available to gather information from job searchers. examines the impact of stemming methods on various natural
Tojan B. Rahhal et.al., (2021) [2] investigated the effects of language processing (NLP) applications and outlines future
participating in Engineering World Health's (EWH's) Design research directions in the field. Yanru Dong et.al,. (2023)
Competition on students' skill development during the [10] introduced a text classification method that enhances
competition and their subsequent careers and surveyed with semantic understanding by leveraging a Self-Interaction
open-ended questions to gather insights from past attention mechanism and dual-label embedding. The paper
competition winners, aiming to assess the potential begins by employing BERT for text feature extraction and
advantages and drawbacks of participation. Naresh Kumar then integrates Self-Interaction attention to capture richer
et.al, (2015) [3] developed a method called Link Context semantics in text representations. Furthermore, the method
Determination (LCD) to ascertain the context of non- focuses on joint labels and word embedding to enhance
descriptive anchor text. It focused on non-descriptive anchor attention learning from text-label pairs. The final
text for this purpose and started by compiling a collection of classification is done using a weighted representation of
online sites from a common domain.These pages were labels. Experimental results demonstrate superior
manually examined to identify relationships between anchor classification accuracy compared to state-of-the-art methods.
text and nearby words. This analysis led to the creation of a S. Jiang, et.al,. (2024) [11] integrated the advantages of
set of rules, organized in a tree structure based on these multiple feature tagging methods and enriched the token
relationships. Walelign Tewabe Sewunetie et.al. (2024) [4] representation. S. Zhang et.al,. (2022) [12] introduced the
introduced novel techniques for parsing phrases within part-aware boundary clue, the edge-guided parsing head is
semantic graph-based induction, employing both ChatGPT- capable of distinguishing adjacent human parts from each
based and Hybrid parser-based approaches. The study other up to 58 parts in a single human instance, even
assesses how well various techniques function in creating overlapping instances. In their 2023 study, Zhang and
semantic networks from text through a comprehensive colleagues introduced a methodology for both mono text and
examination, with a focus on collecting complex event bitext parsing tasks, conducting experiments on Chinese and
descriptions and relationships. Carson Harmon et.al. (2020) English datasets for the former task. The proposed framework
[5] introduced a new approach for labeling the semantic by K. Tu et.al,. (2014) [13] aims to achieve deep semantic
purpose of the functions in a parser. The input file containing parsing of video and text beyond conventional bag-of-words
a predetermined syntax tree is provided to a modified version methods, to conduct parsing and reasoning across spatial,
of the target parser equipped with universal taint tracking. temporal, and causal dimensions using joint S/T/C-AOG
Alexander Grushin et.al. (2022) [6] introduced RL-GRIT, representation, and to demonstrate how deep joint parsing
a reinforcement learning approach for training neural enhances tasks such as generating narrative text descriptions
networks to parse sentences in unknown formats without and answering queries in various forms. M. Shen et.al, (2014)
explicit grammar. The method focuses on learning parsing [14] demonstrated the effectiveness of the approach in
actions to maximize successful parses in a corpus. To address experiments conducted on the Penn Treebank and the
the network's limitation in detecting anomalous sentences, Chinese Treebank. S. Zhang et.al,. (2022) [15] introduced a
the paper presents techniques for rule extraction and anomaly refinement head integrating box-level score and part-level
detection, showing effectiveness in non-regular and high- parsing quality is exploited to improve the quality of the
randomness formats and proposed a two-pass inference parsing results M. Zhang et.al,. (2018) [16] proposed an
method for generating concise rules in high-randomness approach to the mono text and bitext parsing tasks. S. Jaf et
formats, offering insights into common errors in practical al. (2019) [17] developed a parser utilizing transfer learning
formats. Mohammed Kutbi et.al., (2023) [7] developed a techniques within this architecture to effectively tackle
preprocessing methodology which employs a Named Entities crucial challenges associated with under-resourced
Recognizer (NER) to identify named entities, subsequently languages.
substituting them with their respective type such as location, The lack of an integrated framework that combines
person, or organization name, enhancing accuracy and advanced natural language processing (NLP) techniques with
safeguarding privacy rather than eliminating them or automated career path optimization and re-education
allowing them to distort our data. Alexander Grushin et.al. recommendations, customized to diverse professional fields
(2023) [8] conducted a survey on the evolution of Natural other than software engineering, by addressing both technical
Language Processing (NLP), focusing on text representation and contextual nuances of various job markets. This gap
techniques from the 1970s to the present. It discusses various indicates that much of the present research is in specific
embedding types, and their representations address issues, domains or particular aspects of NLP and career guidance,
limitations, and applications, highlighting the NLP field's and thus the findings from those are not generalizable, which
progression in comprehending syntactic, semantic, and shows the possibility of having a more interdisciplinary
sentimental aspects of text. Abdul Jabbar et.al, (2023) [9] approach.

979-8-3315-4066-1/24/$31.00 ©2024 IEEE 1585

Authorized licensed use limited to: Gannon University. Downloaded on December 13,2024 at 04:10:19 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Second International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI 2024)
IEEE Xplore Part Number : CFP24VG0-ART ; ISBN : 979-8-3315-4066-1

III. PROPOSED SYSTEM the system generates a "Resume in standard format" that is
Resspar is a web application designed to streamline the ready for the conversion process. The standard format
resume parsing process using AI techniques. It allows users ensures that the resume is easy to use and understand,
to upload their resumes in PDF format and extracts relevant regardless of the format. Finally, the system "Stops" the
information using tools like GenAI, LLM. Once the conversion process. The system then moves on to the next
information is extracted, Resspar analyzes the skills and resume in the list, and the steps are repeated until all the
filters them by requirements. The aim of Resspar is to resumes in the list are converted. The process ensures that the
simplify the hiring process. resumes are in a standard format that is easy to use and
understand, regardless of the format. The Resume Parsing
System is a highly efficient architecture designed to match
resumes with job descriptions accurately and quickly. It is
made up of several distinct components, each of which plays
a critical role in the system's overall functionality. The Master
DB is the central database that stores all necessary data,
including resumes, candidate metadata, job descriptions, skill
information, personal information, educational information,
and duplicate information. This database is the foundation of
the entire system, and it is constantly updated with new data,
ensuring that the system always has the most up-to-date
information. The Batch Processor is a critical component of
the system, as it processes large volumes of data in batches.
It extracts resumes and job descriptions from the Master
DB and passes them to the Text Index Generator. The Batch
Processor is designed to handle large volumes of data
efficiently and accurately, ensuring that the system can
process data quickly and without errors. The Text Index
Generator is responsible for creating a text index for each
resume and job description. This allows the system to search
and match relevant resumes and job descriptions based on
their content. The Text Index Generator is a critical
component of the system, as it enables the system to quickly
and accurately find relevant resumes and job descriptions,
even when dealing with large volumes of data. The Text
Fig. 1. Architecture of the proposed work
Index is a crucial component of the system, as it is used by
the Query Processor to find relevant resumes and job
In Fig 1. the conversion process begins with the "Start" descriptions. The Query Processor is responsible for
component, which signifies the beginning of the resume processing queries from users and returning the results to the
conversion process. The first step in the process is to "Pick user. The Query Processor is designed to handle a variety of
next resume" from the list of resumes that need to be queries and to return the most relevant results possible. The
converted. By setting the operation type to "Yes," the system Extracted DB is a database that stores the extracted resumes
signals that the conversion process is ready to proceed. Once and job descriptions from the Batch Processor. The Extracted
the system has determined that the operation type is "Yes," DB is then used by the Resume Miner to extract relevant
the next step is to determine the format of the resume. The information, such as skills, education, and personal
system checks if the format is one of the following: "doc," information. The Resume Miner is a critical component of the
"pdf," or "text." If the format is not one of the three supported system, as it enables the system to provide detailed
formats, the system sets the "Operation Type (OT)" to "No," information about each candidate and job description,
indicating that the conversion process cannot proceed. making it easier for users to find the best candidates for a job
However, if the format is supported, the system moves on to or the best job for a candidate. The Duplicate Detector is
the next step. In the next step, the system extracts the contents another important component of the system, as it identifies
of the resume and proceeds to parse each word in the file and marks duplicate resumes, ensuring that the user is
individually. The system checks each word to determine if it presented with only unique and relevant results. This is a
is "Relevant" to the conversion process. If the word is not critical function of the system, as it ensures that users are not
relevant, the system moves on to the next word. However, if presented with the same resume multiple times, which can be
the word is relevant, the system proceeds to the next step. The frustrating and time-consuming. The Results are the resumes
"Generate Key-Value pair" step involves creating a unique and job descriptions that match the user's query, and are
identifier (key) for each relevant word and associating it with presented to the user in a clear and concise manner. The
the word (value) as a pair. This step helps to ensure that the Results are designed to be easy to understand, and provide
relevant words are organized and easily identifiable. The key- users with the information need to make informed decisions
value pairs are then sorted based on the key, which ensures about candidates and job descriptions. The Resume Matcher
that are in a consistent order. Once the key-value pairs are is the final component of the system, and it is responsible for
sorted, the system proceeds to the "Reduce Key-Value pair" matching resumes with job descriptions based on the
step. The system uses a set of rules and algorithms to reduce extracted information and the text index. The Resume
the key-value pairs and convert them into a format that is easy Matcher interfaces with the Master DB, Text Index, Resume
to use and understand. Once the key-value pairs are reduced, Miner, and Duplicate Detector, and by using these

979-8-3315-4066-1/24/$31.00 ©2024 IEEE 1586

Authorized licensed use limited to: Gannon University. Downloaded on December 13,2024 at 04:10:19 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Second International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI 2024)
IEEE Xplore Part Number : CFP24VG0-ART ; ISBN : 979-8-3315-4066-1

components, it can help ensure that the user is presented with the candidate's name, email address, phone number, and skills
the most relevant and accurate results. The Resume Matcher from the resume. These extracted data points are then
is the critical component of the system, as it enables the structured and displayed in a tabular format on the Upload
system to provide users with accurate and relevant matches page, showcasing columns for name, email, phone number,
between resumes and job descriptions. and skills. Fig 3 shows the tabular presentation makes it
convenient for recruiters and hiring managers to review and
IV. RESULTS AND DISCUSSION compare candidate profiles efficiently. Additionally, the
The user authentication system implemented in the parsed data is securely stored in the SQLite database,
Resspar System follows a straightforward logic. In the ensuring that the information remains accessible for future
registration page, when a user enters their desired username, reference and filtering operations. This implementation logic
email, and password and submits the form, the backend code ensures that the resume parsing feature in Resspar enhances
validates the inputs and securely stores them in the "users" the recruitment process by automating data extraction and
table of the SQLite database. This process involves hashing providing a user-friendly interface for candidate evaluation.
the password for security. On the login page, when a user
enters their email and password and submits the form, the
backend code retrieves the corresponding record from the
database based on the provided email. It then compares the
hashed password stored in the database with the hashed
version of the password entered by the user. If the credentials
match, the user is logged in and redirected to the Resspar
home page; otherwise, an error message is displayed
indicating invalid credentials. This approach ensures that user
data is securely managed and authenticated during the
registration and login processes. The user interface (UI)
implementation in Resspar is designed for ease of use and
accessibility. The login and register pages are the entry points
Fig. 3. Filtering Candidates knowledgeable in SQL
for users, where can input their email and password for
authentication. Upon successful login, users are directed to Users are provided with an input section where can
the home page, which serves as the central hub of specify job roles or domains, as well as search for specific
information. Fig 2 shows the home page prominently displays skills required for a particular job. Fig 4 shows the system
basic information about Resspar and its features, including a intelligently matches the specified criteria with candidate
brief description of the resume parsing functionality. At the skills extracted from resumes during parsing. This matching
top right corner of the page, tabs for Upload, Resources, process involves leveraging AI algorithms, NLP techniques,
About, and Contact are provided for easy navigation. The and database querying to identify candidates whose skills
Upload tab leads users to the resume upload functionality, align closely with the given criteria. The filtering logic
where can select and upload their resumes in PDF format. ensures that recruiters and HR professionals can quickly and
The main section of the home page also contains buttons that
accurately identify qualified candidates based on their
direct users to the upload page for convenience. Additionally,
below the main section, users can access Resources for qualifications and suitability for the job roles. This feature
further information, learn more About Resspar, and find empowers recruiting teams to find the most suitable
Contact details for support or inquiries. This UI logic ensures candidates efficiently, promoting fairness in the hiring
a seamless and intuitive user experience, guiding users process by objectively evaluating candidates' qualifications
through key features and information while maintaining a against job requirements. By integrating this powerful
clean and organized layout for efficient navigation. filtering capability, Resspar aims to revolutionize recruitment
practices and contribute to building stronger, more diverse
teams for organizations.

Fig. 2. Resume uploading page

The resume parsing functionality in Resspar follows a

systematic approach. When a user uploads a resume in PDF
format through the Upload page, the backend code processes Fig. 4. Filtering Candidates knowledgeable in C++
the uploaded file. Using libraries like PyMuPDF (fitz) for
document processing and Google GenerativeAI for AI-driven Fig 5 shows the database management in Resspar
parsing, the system intelligently extracts key details such as leverages SQLite for efficient storage and retrieval of parsed

979-8-3315-4066-1/24/$31.00 ©2024 IEEE 1587

Authorized licensed use limited to: Gannon University. Downloaded on December 13,2024 at 04:10:19 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Second International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI 2024)
IEEE Xplore Part Number : CFP24VG0-ART ; ISBN : 979-8-3315-4066-1

resume data. The SQLite database is utilized due to its accuracy and efficiency. Moreover, the system's structured
lightweight nature and suitability for single-user applications presentation of parsed data in a tabular format has greatly
like Resspar. enhanced candidate evaluation, making it easier for recruiting
teams to review and compare profiles based on specific job
requirements and skillsets. The user-friendly interface of
Resspar ensures a seamless experience for all users,
providing intuitive navigation and comprehensive
information about resume parsing, features, resources, and
contact details. Additionally, secure data management
practices, including the use of SQLite and password hashing
techniques, ensure the confidentiality and integrity of user
authentication data and parsed resume information. Looking
ahead, the system's scalability allows for future
enhancements such as integration with external APIs,
advanced AI models, and data analytics, positioning Resspar
Fig. 5. RESUMES table to store data from the parsed resumes as an innovative solution poised to revolutionize recruitment
practices and contribute to building stronger, more diverse
Two main tables are created within the SQLite database: teams within organizations. Overall, the successful
one for managing user authentication information and implementation and achieved results of the Resspar System
another for storing parsed resume data. Fig 6 shows the demonstrate its potential to significantly impact recruitment
"users" table is responsible for storing user credentials processes, improve candidate selection outcomes, and enable
securely. It includes columns such as name, email, and a more efficient and fair hiring environment.
password. When a user registers on Resspar, their credentials
are hashed for security and then inserted into this table. REFERENCES
During the login process, the user's input is compared with [1] S. Ashrafi, B. Majidi, E. Akhtarkavan and S. H. R. Hajiagha,
the hashed password stored in this table for authentication. "Efficient Resume-Based Re-Education for Career Recommendation
On the other hand, the "resumes" table is designed to store in Rapidly Evolving Job Markets," in IEEE Access, vol. 11, pp.
124350-124367, 2023, doi: 10.1109/ACCESS.2023.3329576.
the parsed data extracted from resumes uploaded by users.
[2] J. Baker and T. B. Rahhal, "Fostering Career Exploration and a Global
This table typically includes columns like id, name, email, Perspective," in IEEE Potentials, vol. 40, no. 4, pp. 6-11, July-Aug.
phone number, and skills. When a user uploads a resume, the 2021, doi: 10.1109/MPOT.2021.3067274.
backend processing logic extracts relevant information from [3] N. Kumar and M. Singh, "Stanford parser-based approach for
the PDF file and inserts it into this table. This parsed data is extraction of Link- Context from non-descriptive Anchor-Text,"
then available for filtering and candidate selection purposes Proceedings of 3rd International Conference on Reliability, Infocom
Technologies and Optimization, Noida, India, 2014, pp. 1-6, doi:
within Resspar. SQLite's simplicity, efficiency, and 10.1109/ICRITO.2014.7014751.
compatibility make it an ideal choice for managing user [4] W. T. Sewunetie and L. Kovács, "Exploring Sentence Parsing:
authentication and storing parsed resume data in Resspar, OpenAI API-Based and Hybrid Parser-Based Approaches," in IEEE
ensuring smooth and secure database operations. Access, vol. 12, pp. 38801-38815, 2024, doi:
10.1109/ACCESS.2024.3360480.
[5] C. Harmon, B. Larsen and E. A. Sultanik, "Toward Automated
Grammar Extraction via Semantic Labeling of Parser
Implementations," 2020 IEEE Security and Privacy Workshops
(SPW), San Francisco, CA, USA, 2020, pp. 276-283, doi:
10.1109/SPW50608.2020.00061.
[6] A. Grushin and W. Woods, "Anomaly Detection with Neural Parsers
That Never Reject," 2022 IEEE Security and Privacy Workshops
(SPW), San Francisco, CA, USA, 2022, pp. 88-97, doi:
10.1109/SPW54247.2022.9833865.
[7] M. Kutbi, "Named Entity Recognition Utilized to Enhance Text
Classification While Preserving Privacy," in IEEE Access, vol. 11, pp.
117576-117581, 2023, doi: 10.1109/ACCESS.2023.3325895.
[8] R. Patil, S. Boit, V. Gudivada and J. Nandigam, "A Survey of Text
Representation and Embedding Techniques in NLP," in IEEE Access,
vol. 11, pp. 36120-36146, 2023, doi:
Fig. 6. USERS table to store data from the user authentication 10.1109/ACCESS.2023.3266377.
[9] A. Jabbar, S. Iqbal, M. I. Tamimy, A. Rehman, S. A. Bahaj, and T.
V. CONCLUSION AND FUTURE WORK Saba, "An Analytical Analysis of Text Stemming Methodologies in
Information Retrieval and Natural Language Processing Systems," in
IEEE Access, vol. 11, pp. 133681-133702, 2023, doi:
The implementation of the Resspar System has yielded 10.1109/ACCESS.2023.3332710.
significant benefits and positive outcomes for recruiters and [10] Y. Dong, P. Liu, Z. Zhu, Q. Wang, and Q. Zhang, "A Fusion Model-
HR professionals. Firstly, it has drastically reduced manual Based Label Embedding and Self-Interaction Attention for Text
effort and time spent on resume review and data extraction Classification," in IEEE Access, vol. 8, pp. 30548-30559, 2020, doi:
through its automated parsing capabilities. By streamlining 10.1109/ACCESS.2019.2954985.
the candidate selection process with AI-driven filtering, [11] S. Jiang, Z. Li, H. Zhao, and W. Ding, "Entity-Relation Extraction as
Full Shallow Semantic Dependency Parsing," in IEEE/ACM
recruiters can now swiftly identify qualified candidates with

979-8-3315-4066-1/24/$31.00 ©2024 IEEE 1588

Authorized licensed use limited to: Gannon University. Downloaded on December 13,2024 at 04:10:19 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Second International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI 2024)
IEEE Xplore Part Number : CFP24VG0-ART ; ISBN : 979-8-3315-4066-1

Transactions on Audio, Speech, and Language Processing, vol. 32, pp.

1088-1099, 2024, doi: 10.1109/TASLP.2024.3350905.
[12] S. Zhang, X. Cao, G. -J. Qi, Z. Song and J. Zhou, "AIParsing: Anchor-
Free Instance-Level Human Parsing," in IEEE Transactions on Image
Processing, vol. 31, pp. 5599-5612, 2022, doi:
10.1109/TIP.2022.3192989.
[13] K. Tu, M. Meng, M. W. Lee, T. E. Choe and S. -C. Zhu (2014), "Joint
Video and Text Parsing for Understanding Events and Answering
Queries," in IEEE MultiMedia, vol. 21, no. 2, pp. 42-70, Apr.-June
2014, doi: 10.1109/MMUL.2014.29.
[14] M. Shen, D. Kawahara, and S. Kurohashi, "Dependency Parse
Reranking with Rich Subtree Features," in IEEE/ACM Transactions
on Audio, Speech, and Language Processing, vol. 22, no. 7, pp. 1208-
1218, July 2014, doi: 10.1109/TASLP.2014.2327295.
[15] S. Zhang, X. Cao, G. -J. Qi, Z. Song and J. Zhou, "AIParsing: Anchor-
Free Instance-Level Human Parsing," in IEEE Transactions on Image
Processing, vol. 31, pp. 5599-5612, 2022, doi:
10.1109/TIP.2022.3192989.
[16] M. Zhang, W. Chen, X. Duan and R. Zhang (2013), "Improving
Graph-Based Dependency Parsing Models with Dependency
Language Models," in IEEE Transactions on Audio, Speech, and
Language Processing, vol. 21, no. 11, pp. 2313-2323, Nov. 2013, doi:
10.1109/TASL.2013.2273715.
[17] S. Jaf and C. Calder (2019), "Deep Learning for Natural Language
Parsing," in IEEE Access, vol. 7, pp. 131363-131373, doi:
10.1109/ACCESS.2019.2939687.

979-8-3315-4066-1/24/$31.00 ©2024 IEEE 1589

Authorized licensed use limited to: Gannon University. Downloaded on December 13,2024 at 04:10:19 UTC from IEEE Xplore. Restrictions apply.