The American Journal of Applied Sciences
18
https://www.theamericanjournals.com/index.php/tajas
TYPE
Original Research
PAGE NO.
18-24
10.37547/tajas/Volume07Issue02-04
OPEN ACCESS
SUBMITED
16 December 2024
ACCEPTED
18 January 2025
PUBLISHED
20 February 2025
VOLUME
Vol.07 Issue02 2025
CITATION
Jasti Manohar Sai. (2025). An approach to developing the scafwording
application for vocabulary expansion in gre preparation. The American
Journal of Applied Sciences, 7(02), 18
–
24.
https://doi.org/10.37547/tajas/Volume07Issue02-04
COPYRIGHT
© 2025 Original content from this work may be used under the terms
of the creative commons attributes 4.0 License.
An approach to
developing the
scafwording application
for vocabulary expansion
in GRE preparation
Jasti Manohar Sai
Software Development Engineer at Workday, Atlanta, Georgia, USA
Abstract:
This article examines the process of
developing the Scafwording application aimed at
enhancing vocabulary skills among users preparing for
the GRE exam. The platform employs adaptive
methods that integrate reinforcement learning
algorithms to implement personalized approaches to
the learning process.
The study's objective was to create a tool that
facilitates the acquisition of vocabulary necessary for
completing GRE tasks. The methodology is based on
cognitive principles, including spaced repetition and
simulation of conditions close to a natural language
environment. The technological process of analyzing
user interactions on the platform helps tailor
educational programs to individual needs.
Interactive components and progress visualization,
help maintain engagement in the learning process. The
platform creates an environment for studying words in
context, enhancing their practical application.
The results demonstrate that the Scafwording
application is an effective tool for learning. The
information presented in this study is valuable for
students, educators, and professionals as it offers
opportunities to learn language based on individually
tailored strategies.
Keywords:
Scafwording,
GRE,
vocabulary,
reinforcement learning, spaced repetition, contextual
learning.
Introduction:
Modern educational technologies are
driving the evolution of learning approaches, which are
essential for exam preparation, which requires
The American Journal of Applied Sciences
19
https://www.theamericanjournals.com/index.php/tajas
The American Journal of Applied Sciences
mastering vocabulary and analyzing presented texts.
Traditional methods, such as paper flashcards or
printed study materials, are gradually losing relevance.
They are being replaced by mobile applications
designed to optimize the learning process. However,
most of these programs fail to account for individual
learner characteristics, offering a standardized
material acquisition path that is not adapted to varying
knowledge levels, learning speeds, or preferred
formats of content delivery. A lack of contextual tasks
and limited interactivity further complicates the
practical application of new words in real-life
scenarios.
The relevance of developing educational tools that
cater to individual user needs is supported by
advancements in cognitive science. Technologies
integrating adaptive algorithms enable the creation of
solutions tailored to the unique characteristics of each
learner. An example is an application employing
algorithms that develop personalized learning
trajectories.
This study aims to create a conceptual model of an
application
designed
to
enhance
vocabulary
acquisition during exam preparation. Emphasis is
placed on personalized learning, leveraging data to
adapt the educational process, and analyzing
opportunities to extend the tool's application to other
areas.
METHODS
Scientific studies dedicated to mobile applications for
learning English and preparing for exams encompass a
wide range of approaches. Researchers address
aspects such as interface design, pedagogical methods,
and evaluation systems for software.
Articles focusing on application development
emphasize usability, adaptation to individual needs,
and the creation of intuitive solutions. The work by
Martell M. et al. [1] describes interfaces tailored for
exam preparation, designed to meet the needs of users
learning English as a foreign language. The study by
Kazemainy F. et al. [2] examines the process of
developing software that adapts to various
educational tasks. Research by Martínez R. F. et al. [3]
highlights the use of spaced repetition algorithms for
memorizing terms.
The reviewed studies focus on methodologies
leveraging
modern
technologies
to
achieve
educational goals. Articles by Al-Jarf R. S. [4,5] analyze
the practical value of mobile applications and
flashcards for exam preparation. Aslan M. and Tütüniş
B. [6] explore the personalization of learning processes
through mobile devices. Research by Mao Y., Mofreh
S. A. M., and Salem S. [7] investigates the impact of
smartphones on educational processes. The work by
Ene A., and Stirbu C. [8] describes tools designed to
expand vocabulary.
Scientific studies in this field concentrate on
developing
methodologies
for
analyzing
the
functionality of educational software. The article by Lin
C. H. et al. [9] proposes a framework for objectively
evaluating mobile applications. In the study by Mirzaei
S. et al. [10], the VLASTWA platform is analyzed, which
integrates vocabulary learning with the development
of strategic thinking skills.
In turn, the source [11] demonstrates a web
application posted on the official website for
expanding vocabulary in preparation for the GRE.
The work described in [12] examines the
implementation of reinforcement learning methods in
a system designed to personalize the educational
process. It emphasizes that the use of a model based
on AI algorithms allows for the adaptation of actions
depending on a learner's achievements or errors,
thereby facilitating modifications to teaching methods.
Such a system adjusts the learning strategy based on
results, ensuring the effective assimilation of material.
It is argued that its implementation will improve
outcome forecasting, enabling the system to adapt the
process to each learner.
Thus, the scientific literature reflects the diversity of
approaches. Some authors focus on creating user-
friendly interfaces tailored to learners, while others
emphasize the educational aspects of technology use.
Limited attention has been given to tools analyzing the
development of oral communication skills. Issues
related to the universality of applications for different
cultural and linguistic groups require further
investigation.
A
promising
direction
involves
developing methods that integrate interdisciplinary
approaches to enhance the effectiveness of
educational platforms.
An analysis of existing commercial and non-
commercial applications for exam preparation reveals
that many solutions while incorporating individual
elements of theoretical frameworks, fail to fully realize
the potential of modern adaptive approaches. Popular
vocabulary memorization applications such as Quizlet
and Memrise provide basic functionality, including
user-generated
flashcards,
tests,
and
some
gamification elements. More specialized platforms
designed for GRE preparation (e.g., Magoosh, Kaplan,
Barron’s) offer predefined vocabulary lists and practice
exercises. However, in most cases, adaptation to
individual learner characteristics is limited to adjusting
difficulty levels or repeating incorrectly answered
The American Journal of Applied Sciences
20
https://www.theamericanjournals.com/index.php/tajas
The American Journal of Applied Sciences
questions, adhering to a static model. Features such as
multiple iterations with varying levels of detailed hints
and interactive feedback to strengthen semantic
associations between a word and its meaning are often
absent.
An evaluation of current solutions indicates that the
methods employed often fail to adequately
incorporate recent advancements in machine learning
and behavioral data analysis. Notably, approaches
based on reinforcement learning and intelligent
decision-making models in educational environments
remain
underutilized.
This
situation
appears
paradoxical given the extensive empirical data on user
experience and learning outcomes. Additionally, many
applications offer only a limited range of contexts for
vocabulary presentation. Research suggests that
presenting vocabulary through multifaceted contexts,
such as dialogues and stories, enhances retention.
However, many platforms rely primarily on isolated
definitions or sentence fragments.
When comparing existing solutions, some platforms
implement basic adaptive mechanisms. For instance,
Magoosh periodically revises the set of words
presented based on test performance, and Quizlet
allows users to create custom term sets. However,
these adaptations are often simplistic and lack flexible
adjustments to tailor the learning trajectory to
individual cognitive models. Consequently, users
encounter challenges such as inefficient repetition of
words, suboptimal increases in complexity, and
insufficient attention to previously overlooked terms.
The absence of dynamic difficulty management,
limited contextual cues, and ineffective integration of
user behavioral patterns restrict the potential of
current systems.
Thus, a review of the literature and existing
applications demonstrates that despite the availability
of theoretically grounded strategies (e.g., spaced
repetition, contextualization, gamification), the
market is dominated by tools that fail to fully leverage
modern
intelligent
approaches
to
learning
personalization and adaptation. This creates an
opportunity for the development of a new solution
capable of addressing the limitations of current
systems. In the practical section of this study, a
conceptual framework for an application will be
presented. This application will integrate data analysis
methods, reinforcement learning algorithms, and
enhanced vocabulary contextualization to create an
individualized trajectory for each user, taking into
account their proficiency level, information processing
style, and learning progress dynamics.
RESULTS AND DISCUSSION
The Scafwording application is based on findings from
scientific studies examining memory and information
retention processes. Its functionality adapts to
individual user needs, beginning with assessing the
user's proficiency level, customizing word repetition
frequency, and generating personalized word lists. This
includes a system of achievements and daily tasks.
The
contextual
learning
feature
facilitates
comprehension of words through examples in
sentences and exam-style tasks, including fill-in-the-
blank exercises and analogy challenges.
The application's software component analyzes user
learning activity to identify optimal approaches,
adjusting the focus on words that present greater
difficulty. This enables a tailored learning process
aligned with the user's proficiency level and goals [1,
2]. The interface is designed for simplicity and
convenience, allowing users to focus on studying.
Progress is visualized through clear graphs displaying
the number of words learned and retention
percentage.
Additionally,
users
can
access
consultations and recommendations to enhance
learning efficiency.
The development of the application is guided by three
principles: flexibility, interaction, and engagement.
These principles ensure a personalized learning
process, sustain user attention, and maintain interest
in the learning activities [3, 6]. Table 1 below outlines
the principles of application development.
Table 1. Principles of Application Development [4, 5, 7, 8]
Principle
Description
Flexibility
The application uses algorithms to process user data, accounting for individual
challenges. Material that poses difficulties is repeated, while mastered content
is excluded from further tasks.
The American Journal of Applied Sciences
21
https://www.theamericanjournals.com/index.php/tajas
The American Journal of Applied Sciences
User
Interaction
Gamified elements such as quizzes make the learning process more engaging.
Progress is demonstrated through visual graphs and tables.
Engagement
The application includes daily tasks, enabling the creation of communities
focused on collaborative learning.
From a technological perspective, the Scafwording
application utilizes cross-platform methods, ensuring
accessibility for users of all devices. Data analysis is
conducted using algorithms that generate predictions
to optimize the sequence of material presentation. The
simplicity of the interface is achieved through
minimalism, an intuitive structure, and adaptive design
elements [6, 9]. Figure 2 below illustrates the sequence
of questions used in the daily quiz.
Fig.2. The sequence of questions used in the daily quiz
Vocabulary within the application is selected based on
an analysis of frequency usage in exam-related texts.
Thematic collections are created to cover various
fields. Each word is accompanied by:
●
a detailed description and usage examples,
●
a list of synonymous and antonymous
meanings,
●
audio recordings by native speakers,
●
visual elements that create associative
imagery.
The vocabulary learning tools implemented in
Scafwording include a spaced repetition system to
reinforce the material, exercises incorporating words
into texts, sentences, and paragraphs, and session
results that provide knowledge assessments and
recommendations for further study.
The application design focuses on developing
mechanisms that adjust task complexity based on user
capabilities. A flexible difficulty adjustment system is
created, and materials are adapted to regional
educational characteristics.
The application is versatile and can be used in various
domains. Students gain a tool tailored to their learning
style and proficiency level. Educators can track learner
progress and refine their teaching methods.
Researchers in the field of language acquisition use the
platform to gather data on information retention
mechanisms.
The functionality extends beyond exam preparation.
The program is suitable for language learning,
mastering specialized terminology, and applying
gamified educational methods. Personalization
enables the platform to be tailored to specific
objectives, providing materials aligned with user goals
[4,5,10].
The sequence of questions that
are used in the daily quiz
The beginning of the quiz
On the control panel, the user
selects "Start a Daily Quiz".
The quiz consists of words that
have been marked for
verification.
Word representation
The user is offered a word (for
example, "abate"), and he
must choose the correct
meaning from four possible
options. No hints are provided
during the test.
Completing the test
after the quiz the results are
shown and yes memorization
score is updated.
The American Journal of Applied Sciences
22
https://www.theamericanjournals.com/index.php/tajas
The American Journal of Applied Sciences
Example of a user interface for the learning flow:
1.
Present one word at a time, offering the
options "Know" and "Don't Know."
2.
When the user selects "Don't Know," display
hints (context, dialogue, or story) one at a time.
3.
After showing a hint, present a question with
multiple-choice options as easily clickable buttons with
clear feedback.
4.
For correct responses, highlight the option in
green, proceed to the next word, and repeat from step
1.
5.
For incorrect responses, highlight the option in
red, display another hint, and disable it.
6.
Once hints are exhausted, show a brief
definition of the word and move to the next word,
restarting from step 1 until the session words are
completed.
7.
At the end of the session, display a summary
of learned words, marking incorrect words for further
review.
The user interface for the daily quiz includes:
●
Displaying each word with four answer choices
below it.
●
A summary of results at the end of the quiz.
This summary should include an accuracy scale and
personalized feedback based on accuracy.
●
The word review section should display words,
results, and an option to review each word. When the
user selects "Review Word," a screen appears showing
the correct and selected answers. After completing the
individual review, the user returns to the word review
screen [6, 7, 9].
The implementation of the hint mechanism was
achieved using AI methods such as SARSA and Q-
LEARNING. These algorithms are designed to operate
under uncertainty and rely on previous experiences.
SARSA is a method in which learning is built within the
framework of its policy. The algorithm uses past
actions and their consequences to adjust future steps.
Q-LEARNING, on the other hand, employs off-policy
learning. This method allows the algorithm to explore
various strategies, taking into account long-term
rewards, and enabling it to act independently of prior
decisions. For example, consider an employee who has
just joined a team. If their training is based on
experience gained through their strategy, this
illustrates policy-based learning. Conversely, if training
occurs through observing colleague
s’ actions and
applying that knowledge to their work, it represents
off-policy learning.
The
anatomical
simulator
model
presented
demonstrates the operation of an intelligent tutoring
system (ITS) developed according to general ITS
principles. In this system, the student interacts with a
web interface connected to a tutor model that
substitutes for a human instructor. The tutor generates
questions based on various scenarios stored in the
content database. These questions are then sent to the
student, who responds.
The tutor evaluates the accuracy of the responses and
saves the progress information in the system
supporting the student model. The results are
accumulated in a centralized database. Depending on
the responses received, the tutor may offer additional
hints, ask new questions, or modify the scenario,
allowing the learning process to be tailored to the
current needs of the student.
The development process of the simulator for the
Interactive Training System involves the use of
scenario and student classes, which facilitate data
extraction and display during the training process.
Artificial intelligence interacts with students according
to predefined schemes. Responses are analyzed based
on a conceptual map: correct answers lead the system
to proceed to the next question, while incorrect ones
prompt the provision of a hint. At the initial stage, the
probability of receiving a hint is 33%.
The simulator's structure is organized into three cycles:
the student cycle, the scenario cycle, and the question
cycle. The question cycle ends when all questions for a
given scenario are completed. The scenario cycle
concludes when the student has answered all
questions across all scenarios. The training cycle ends
once all students have completed all assignments.
During interactions between students and the artificial
tutor system, each student’s data is saved for analysis.
The functions that process this data are described in
Table 2.
Table 2. Description of data processing functions [12]
Feature name
Description
Time stamp
Timestamp is stored when the student
has entered his/her answer
The American Journal of Applied Sciences
23
https://www.theamericanjournals.com/index.php/tajas
The American Journal of Applied Sciences
Student
Name of the student
Scenario description
The description of a scenario where
the student is answering a question
Question
The question which the student is
solving
Student question thinking time
End of the tutor’s question to the
beginning of the student’s answer or
the End of the tutor’s hint to the
beginning of the student’s answer
Student answer
The answer that the student has given
for a question under a scenario.
Correct?
This feature indicates whether an
answer is right or wrong
Hint
The hint number
Hint type
The type of hint; whether a hint given
is in video, text, or diagrammatic
format
Hint time
The time needed by an average
student to read/view a hint
The reinforcement learning method enables the
modeling of scenarios where both the sequence of
actions and the choice itself are critical. The application
of these algorithms contributes to the optimization of
the entire process. For instance, after several video
hints, a brief textual or schematic hint summarizing the
material can prove effective, despite video hints
generally demonstrating better results. To determine
the optimal strategy for selecting hints, the type of hint
was randomly chosen whenever a student was unable
to provide the correct answer. Each type of hint was
selected with an equal probability of 131/3%. After
1000 iterations, an analysis of the data revealed the
distribution of hint types.
A similar approach with epsilon reduction was used to
develop personalized recommendations tailored to
each student. The objective was to compare the
effectiveness of Q-learning and SARSA in optimizing
hints adapted to individual learner needs. Q-learning
demonstrated the highest average reward. This
method updates Q (S, A) values more frequently, which
is particularly evident when observations are limited
[12].
Based on the above, Scafwording can be characterized
as a learning application designed for exam
preparation, aimed at developing skills in text analysis,
context comprehension, and terminology acquisition.
Its functionality is powered by data processing
algorithms that adapt the learning process to the
user’s charact
eristics.
The program records mistakes, collects data on time
spent completing tasks, and analyzes learner behavior.
Based on the gathered information, it builds a model
to create optimal repetition strategies. Tasks are
distributed according to the user’s
mastery level, with
studied topics appearing less frequently. Gamification
elements make the learning process more engaging.
Scafwording integrates modern technologies and
educational methodologies, effectively organizing the
learning process to address linguistic and professional
challenges.
CONCLUSION
The Scafwording application is designed to expand
vocabulary in preparation for the GRE exam. Its
development is based on reinforcement learning
algorithms that utilize user data analysis to create
individualized educational trajectories. This approach
addresses learner-specific characteristics, enabling the
creation of personalized programs.
The application’s functionality encompasses foreign
language
learning,
professional
terminology
acquisition, and integration into corporate training.
This versatility positions the platform as a universal
solution for students, educators, and cognitive activity
professionals.
The concept combines data analysis algorithms with
cognitive psychology methods. Implementing this
The American Journal of Applied Sciences
24
https://www.theamericanjournals.com/index.php/tajas
The American Journal of Applied Sciences
model provides opportunities for the development of
educational programs that meet modern demands,
accommodating the needs of various user categories.
REFERENCES
Martell M. et al. Human-oriented design of
applications for GRE preparation for those for whom
English is not their native language // Proceedings of
the annual meeting of the society "Human Factor and
Ergonomics".
–
Sage, California: Los Angeles, California
: SAGE Publications, 2020.
–
Vol. 64.
–
No. 1.
–
pp.
1095-1099.
Kazemaini F. and others. Flexible development of a
custom dictionary mobile application: a critical
qualitative approach //The quarterly journal "Teaching
English as a Second Language" (formerly "Journal of
Teaching Language Skills").
–
2020.
–
vol. 38.
–
No. 4.
–
pp. 43-93.
Martinez R. F. et al. Development of an application for
studying the academic vocabulary of the English
language using the interval repetition method // 2J.:
Rev. de Tecnol. Calculation.
–
2021.
–
pp. 11-20.
Al-Jarf R. S. Standardized test preparation using mobile
card applications //Online delivery.
–
2021.
–
vol. 3.
–
No. 2.
–
pp. 33-40.
Al-Jarf R. Learning vocabulary in the App Store by EFL
College students // Online broadcast. - 2022.
–
Vol. 5.
–
No. 1.
–
pp. 216-225.
Aslan M., Tyutyunish B. Vocabulary development using
mobile devices and applications // Pearson journal of
social sciences & humanities. - 2022. - Volume 7 (20). -
pp.18-30.
Mao Yu., Mofre S. A.M., Salem S. Effective applications
for smartphones in preparation to the English language
test in college //Teaching English.
–
2024.
–
Vol. 17.
–
No. 9.
–
pp. 123-123.
Ene A., Stirbu S. Java applications for learning English
vocabulary //14th International Conference on
Electronics, Computers and Artificial Intelligence
(ECAI), 2022.
–
IEEE, 2022.
–
pp. 1-4.
Lin K. H. et al. Development of a system for evaluating
applications for learning vocabulary //Interactive
learning environments.
–
2023.
–
vol. 31.
–
No. 10.
–
pp. 7377-7391.
Mirzaei S. et al. Usability testing of VLASTWA web
application for learning vocabulary and strategies
//Proceedings of the 32nd Australian Conference on
Human-Computer Interaction. - 2020.
–
pp. 614-621.
Scafwording.
[Electronic
resource]
Access
mode:https://scafwording.app/.
Jasti Manohar Sai Using Reinforcement Learning in A
Simulated Intelligent Tutoring System. [Electronic
resource]
Access
mode:
https://www.cs.niu.edu/pdf/manohar-ThesisV12.pdf
2019. С. 5
-53.
