See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/377209836

International Journal of Advanced Research and Publications Digital Learning

For Kids

Research · November 2022

DOI: 10.13140/RG.2.2.26112.81924

CITATIONS READS
0 65

4 authors, including:

Kithusshand Raveendran
Sri Lanka Institute of Information Technology
1 PUBLICATION 0 CITATIONS

SEE PROFILE

All content following this page was uploaded by Kithusshand Raveendran on 07 January 2024.

The user has requested enhancement of the downloaded file.

 International Journal of Advanced Research and Publications
ISSN: 2456-9992

Digital Learning For Kids

Sharmilan Sureshwaran, Nilakshana Rajeswaran, Kithusshand Raveendran, Sriram Rajeswaran
Sri Lankan Institute of Information Technology, Department of Information Technology,
Jaffna, Sri Lanka, PH-001 657 567 5676
[email protected]

Sri Lankan Institute of Information Technology, Department of Information Technology,

Batticaloa, Sri Lanka, PH-001 657 567 5676
[email protected]

Sri Lankan Institute of Information Technology, Department of Information Technology,

Kalmunai, Sri Lanka, PH-001 657 567 5676
[email protected]

Sri Lankan Institute of Information Technology, Department of Information Technology,

Jaffna, Sri Lanka, PH-001 657 567 5676
[email protected]

Abstract: The current educational system changed to use the digital learning system due to the Covid-19 pandemic situation. There is a huge
need for the Digital Learning platforms that could use by the students and kids for the educational purpose. The currently available digital
learning platforms needs to enhance, ensure the safety for kids, and can provide variety of services for the kids using the newly available
technologies like Text to speech (TTS), Machine Learning (ML), Image Processing, and Web scraping Technologies. The research paper is
proposing a new system that could help to the kids to the learn the things very easily via watching the videos that is generated automatically
using the new technologies.

Keywords: Text To Speech (TTS), Machine Learning(ML),Image Processing, Web Scraping.

I. INTRODUCTION paper cutting edge technologies like Machine Learning, NLP

The Educational Sector is one of the most affected sectors due (Natural Language Processing), TTS(Text To Speech) and Web
to the covid-19. The Educational system changed a lot with scraping are used. When the user give input keyword the
the Covid-19 pandemic. It caused an unprepared transi- tion to system will extract information and sentiment analysis from
build emergency remote learning platforms. The whole the text and audio input.Then, generating text description using
schools, academy, colleges, and universities are closed and extracted infor- mation and summarizing that text description
moved to the online educational system like zoom, Google into meaningful sentences.Identify complicated terms in
meet also including social media, email, telephone and even meaningful statements and replace them with simple words
post. These online educational systems highly depend on the before translating the sentence to audio. Extracted information
digital learning platforms. Most of the educational organisa- was found by finding relevant photos from web sources, and then
tions started broadcasting video lessons by the television or merging audio and visuals to build and create a digital video
the online distance learning platforms. In this research, A production system. Natural Language Processing is a branch of
Digital Learning Platform for kids is introduced. With the loss linguistics, computer science, and artificial intelligence
of interaction between the teachers and the students, kids are concerned with computer-human interaction, particularly how to
the most effected. They are in the stage of development period design com-puters to process and evaluate huge volumes of
for their life as a human being. They can better absorb content natural lan- guage data.Natural Language Processing frequently
by watching a video than listening to an audio book or involve speech recognition, natural language understanding and
reading a book. Because it is very effective, engaging to see, natural language generation. Machine learning is an area of
and effectively communicates the contents meaning. As a research focused on compre- hending and developing ”learning”
result, the majority of people create videos to explain their processes, or processes that use data to enhance performance on a
information. Today video learning is widely accepted given set of tasks. It is considered to be a component of artificial
learning method. While studying, children may become intelligence.Machine Learning has become a significant
distracted by book reading, audio learning, or other methods. differentiate for many use cases.Machine Learning can be
Kids can’t be as comfortable studying with book reading and categorized by how an algorithm learns to more become accurate
audio learning methods, and they can’t learn whenever and in predictions.The categories are Supervised Learning,
wherever they want. Outsourcing can have an impact on Unsupervised Learning, Semi-supervised Learning and
children’s studies when they are studying using other Reinforcement Learning. The likeness to the human voice and the
methods. Other factors may cause them to lose their ability to clarity with which it can be comprehended determine the quality
study effectively. Video production is expensive when done of TTS, or more specifically, the quality of a speech
manually. Making videos by hand takes far too long. Making synthesizer. The use ofthe program is significantly impacted by
videos using the cutting-edge technologies is aimed. In this the TTS’s lack of emotions in the sound. As a result, it identifies
the primary challenge in the development of TTS, which is the

Volume 5 Issue 11, November 2022 67

www.ijarp.org
 International Journal of Advanced Research and Publications
ISSN: 2456-9992

creation of a sound from text that is as natural-sounding as the to the conception, design and innovationof these systems. [5]
human voice. The goal of TTS is to be able to produce the
entire speech spectrum of a human being, including all speech B. TTS
variations, and to lessen the robotic gasp of the output Existing information and communication technologies, de-
voice in order to increase the difference between human veloped platforms and applications for e-learning allow to
performance and machine voice. The internet is filled with successfully switch to e-learning, if this is a preferred form of
data, thus in order to fully utilize it, its content must undergo education or necessary. The mentioned applications and plat-
specific processing, and a unique online information system forms provide the necessary functions and tools for creating,
must be created. The categorization of online pages is editing and presenting the educational content in electronic
suggested as a crucial stage in those systems. Various machine form. In order to implement successful e-learning,the techno-
learning techniques have used to categorize web pages in this logical competence of teachers must be ensured, which will
research. Web scraping (also known as Screen Scraping, Web ensure the achievement of educational goals. [6] States that
DataEx- traction, and Web Harvesting etc.) It is atechnique in the demands to synthesized speech are more natural,
which meaningful data from the HTML of websitesare personalized and emotional.Speech synthesis technol- ogy
extracted and stored into a central local database based on cloud source and embedded TTS systems would
orspreadsheet. The goal of a Web scraper isconcentrated on provide higher quality and more convenient for users, and they
conversion of unstructured data while preserving in would become the hot spots of applications in the future.With
organized databases. Web scraping is presently cast-off on the development of speech synthesis, text to speech synthesis
various aspectsincluding online price comparison, weather will play a greater role in social life. [7] [8]In this paper,
data monitoring,website change detec- tion, Web mashup, Min-Jae Hwang, Ryuichi Yamamoto; Eunwoo Song, and Jae-
Web research and Webdata integration. Further, it may be Min Kim propose a TTS-driven data augmentation method
noted that Web scrapingmight be alongside the tenures of to improve a quality of non-AR TTS system. Using a large-
usage of few websites. In this research it is introduced a scale synthetic TTS database generated by high-quality AR
system which take a keyword or audio as input and using web TTS systems, they successfully improved the quality of the
scraping related paragraph or images will be extracted from a target TTS System. The future studies should test the
relevant website such as Wikipedia.Then using the extracted augmentation with various phoneme distributions, such as
image and the description a creative video is to be generated. uniformly distributed case. [8]
[1] [2]
C. NLP
II. RELATED RESEARCH Pre-trained models can be used to extract polarity fromdatasets
without the need to build a model from scratch. XLNet
A. Digital Learning outperforms in more tasks compared to other models but its
Zaharah, Indrayanto, C. Dhien Nourwahidah, A. Saehudin, computational complexity is very high. Performance
H. Hasan, and Kamarusdiana states that Information and evaluation with the two latest models, BiGRU and BERT using
Com- munication Technology (ICT) is a means that makes it IMDB dataset shows an improvement in accuracy with number
easier for humans to obtain, process, and communicate of epochs. The study can be further extended by comparing
information without being limited by space and time. As a the performance of various models with more datasets. [9]
result of the covid-19 outbreak, the impact of e-learning has There are five important tasks in Natural Language Pro-
an impact on all sectors, especially in the field of cessing which enable the machine or a computing device to
education. Government policy requires learning to be understand the human language. The accuracy with whichthese
carried out from home, this is to prevent the transmission of tasks are performed determine the level of natural language
the disease. Students are accustomed to learning online and understanding by the machine. This paper showcases how
many lecturers are still not proficient in using information researchers has successfully employed different machine
technology to teach using. [3] The students of early Z learning and deep learning techniques in performing such
generation at these Hungarian universities still like the tasks. These techniques have performed them far better than
traditional type of learning combined with e-learning, that is any of the traditional techniques. [10] Developed language
they like classroom activities and face- to-face education but model, based on deep learning meth- ods, can be used to
at the same time studying on their smart phones. For the generate new sequences of text that have the same statistical
Hungarian universities the application ofblended learning is a properties as the source text. Applicationof relevant language
possible transition in today’s educational shift and tertiary corpus as a training text for deep language model learning can
education should invest in providing e-learners with varying be practically used to improve speech recognition. The results
degrees of computer literacy. [4] Some e-Technologies and presented in this article concern the Polish language but the
especially the adaptive e- Learning can be useful, since it achieved research results can also be applied to other
constitutes the personal tutor for students. There are many languages. [11]
challenges in relation to the appropriate methods and
strategies, which must be applied to meet the needs of D. Web classification
students. This work, which is the result of a brief review of In this work L. Safae, B. E. Habib, and T. Abderrahim review
e-Learning and e-Technology systems, allowed to focus on the existing machine-learning algorithms used for web page
the technological and operational aspects that can contribute classification and compare them to related methods based on
some characteristics. For future work, the visual analysis of web

Volume 5 Issue 11, November 2022 68

www.ijarp.org
 International Journal of Advanced Research and Publications
ISSN: 2456-9992

pages, the removal of noisy content and the implicit and

explicit links with other pages should be taken into
consideration, in order to have the maximum accuracy
possible. [12] This paper tries to solve the problems of
webpage classifi- cation by combining structure and
semantics of. The research trains on the basis of the CNN
model framework. There are still many aspects that need to
be further studied and improved on the problem of web page
classification. Given the chosen sizes of the training corpus
and test corpus have a certain the impact on the
classification. [13] In this paper, they have explored how
well to classify phishing URLs from a set of URLs
containing benign and malicious URLs. they have also
discussed the randomization of the data set, feature
engineering, feature extraction using lexical analysis host-
based features and statistical analysis. In the future, they
would like to incorporate a rule-based prediction based on
the content of a URL in order to provide a comprehensive
solution for phishing URL detection. [14]

E. Web scraping
In this research paper, Stephanie Lunn; Jia Zhu; Monique
Ross show that web scraping and Natural Language
Processing (NLP) can be used to obtain and analyze
information from internet sources. they argue that computer
science educators should consider using these tools to
further their own work and to inform pedagogical practice.
Since new information is constantly being generated, finding
new acquisition methods can only serve to benefit
education. [15] In this paper, V. Singrodia, A. Mitra, and S.
Paul have reviewed the various aspects of Web Scrapper.
they have seen the operating principle, strength and
drawbacks and finally viewed the applications of web
scrapping system. Web scraping is a recognizable phrase A. Extract the information
which has expanded its meaning owing to the requirement While generating the text descriptions for relevant key words
of ”free” data. [16] The main outcomes of this project were web scraping is used.Web scraping is basically extract- ing
user friendly search interface, indexing, query processing, data from website using a bot or a web crawler.It is an
and effective data extraction technique based on web automated process which is using Hypertext Transfer Protocol
structure,form submission analysis and new submission plan. or web browser.First it is used to fetch data from website
Hidden web data need synthetic and semantic matching to and extract data from it.Web page content is downloaded as
fully achieve automatic integration in this thesis fully fetching and stored for later processing.Later they are
automatic and domain dependent prototype system is processed and stored in a database or excel files.
proposed that extract and integrate the data lying behind the
search form. [17] B. Generating text description using extracted
information and summarizing that text description into
III. METHODOLOGY meaningful sen- tences.

There are many libraries can be used for web scraping.Such as

Beautiful Soup, newspaper, also selenium can be used for
automation purpose. Beautiful Soup is a python package.It can
be used forparsing HTML and XML documents.It created parse
tree from the parsed pages, so that They can be used for
extracting data. Selenium is an open source umbrella project
can be used for browser automation. Newspaper is a python
module used for the same purpose.It uses advance algorithms to
extract all the useful text from a website.Newspaper is used in
our project as it extract all the information from a website.As
the text output is aimed for study purpose.

Volume 5 Issue 11, November 2022 69

www.ijarp.org
 International Journal of Advanced Research and Publications
ISSN: 2456-9992

downloaded and tokenized into sentences.Then using SpaCy

detected the tense of the sen-tence , part of speech, dependency
relation and the language.In order to tokenize the text used
WordNetLemmartizer from the NLTK. Lemmartizer is a
process of grouping inflected forms of words together.So that
they can be analysed as a single item.Lemma is based on
intended meaning of the word.Also there is a similar process
called stemming.But lemmatisation identify the efficient
intended part of speech not only within the sentence but also
with the neighbouring sentences and also from the entire
TextBlob is also python library for text processing.it provide a document. Then the entire document was tokenize into
common API for part-of-speech, phrase extraction, sentiment words.In the content analysis all 16 classes was used.
analysis classification and translation.Here TextBlob is used
for analysis.

C. sentiment analysis from the text and audio input.

Sentiment analysis is an information extraction method from
text input.It is basically a text mining technique that uses
Machine Learning and Natural Language Processing, Then NLTK was used for sentiment analysis, SentimentIn-
automatically analyse text for the sentiment of the writer.The tensityAnalyzer from NLTK sentiment package.It takes input
overall purpose is to derive high quality information and as a string and return a dictionary of scores for each category.
actionable insights from text.Aspect based sentiment analysis Speech recognition was used for audio input identification
organise text into categories and then when mining text it purpose.Recognizer class was used by creating an instance to
givesoutput as positive or negative. recognize the audio source. Here the main and the most
important thing is to notify the system to only add positive
content.

D. Website Classification
Classify websites is needed ac-cording to the categories.
Which are

Technology going to use in information extraction from text

and audio Speech recognition is a method to recognize
words spoken aloud and convert into readable text.It first
analyse the audio then breakdown it into parts, digitize it
into computer readable parts and finally use an algorithm to Machine Learning classification algorithms are a supervised
match it to most suitable text representation. Technology learning approach.Main purpose is to classify data in to given
going to use in sentiment analysis from text andaudio classes.Both structured and unstructured data can be
performed.Always in the beginning the class is predicted which
the data points belongs.

Machine Learning algorithm extract the features from the

provided text according to the given tags.Then the classifier It is an approximation.Mapping function will map input variables
model classifies the text into the relevant tags. NLTK is a to discrete output variables.There are many terms used in
Natural Language Processing toolkit which is a suite of classification.Such as Classifier, Classification Model, Feature,
libraries and programs for symbolic and statistical Natural Binary Classification, Multi-class classification, Multi label
Language Processing in English language. SpaCy is a classification, initialize, train the classifier, predict the target,
open-source software library for advanced natu- ral evaluate. There are many classification algorithms used in
language processing, written in the programming languages classifi- cation.Logistic Regression, Naive Bayes Classifier,
Python and Cython. First the particular text input is Stochastic Gradient Descendent, K-Nearest Neighbour, Decision

Volume 5 Issue 11, November 2022 70

www.ijarp.org
 International Journal of Advanced Research and Publications
ISSN: 2456-9992

Tree, Random Forest, Artificial Neural Networks, Support The reached accuracy was,
Vector Machine are some machine learning algorithms used in
clas- sification. Naive Bayes Classifier has been as the main
algorithm and other algorithms.

1) Naive Bayes Classifier: This classifier based on 3) Linear Support Vector: Here the training data is repre-
Bayes’s Theorem.It gives an assumption of independence sented by points in a space. Separate the categories by a gap as
among other predictors.It make sense that when there is a wide as possible.when new points added it has to decide which
particular feature in the class, it is not related to another space does it belongs to. This algorithm is highly effective
feature present in the same class. All properties contribute in high dimensional spaces and memory efficient. But the
independently to the probability even they depend on each algorithm does not directly provide probability estimates.
other.This algorithm is easy to use for comparatively
large data sets.In the below figure it is given the Bayes
Theorem which use to implement Naive Bayes Classifier.

The reached accuracy was,

Confusion Matrix
Confusion matrix is a popular measure for classification
problems.Confusion matrix visualize the performance of a
classification matrix.It represent the counts from predicted
and actual values.This figure shows the formula which uses
to calculate the accuracy of a model. Since accuracy might
be deceptive when applied to unbal- anced data sets,
alternative metrics based on confusion matrix are also
relevant for assessing performance. The reached accuracy was,

4) Decision Tree Regressor: Decision tree build the clas-

sification model as tree structure.This uses if else rules.If else
rules are equally exhaustive and mutually exclusive for
classification.The algorithm break down data into small
structures and the build them assertively incremental decision
tree.The rules are used sequentially and use training data one
at a time.when the rule is learned rules will be removed which
cover tuples.process continue until the termination point
meets.The final structure looks like a tree with nodes and
leaves.

2) Random Forest: Random Forest is a Ensemble

learning method for classification.It construct multitude of
decision trees while constructing.It give output as a
classification or mean prediction of individual trees.It fits
a number of trees on various sub samples of data
sets.Then it uses the average to improve the accuracy of the
model’s predictive nature. The original input size and the
sub sample size is always similar.
Random Forest is more accurate than the decision tree.Because
of the reduction of the over fitting. But Random forest is
complex for implementation and become slow when it
predicts in real time. It is a bad estimator but fast in nature in
comparison to other classifiers.But it needs only small amount
of training data to estimate parameters to get the best results.

Volume 5 Issue 11, November 2022 71

www.ijarp.org
 International Journal of Advanced Research and Publications
ISSN: 2456-9992

forcomputation is pretty high.

The reached accuracy was,

6) Logistic Regression: This classification algorithm use

in machine learning when there is one or more independent to
de- termine an outcome.Always it has two possible
outcomes.The main goal of the algorithm is to make best
fitting relationship between the the set of independent
variables and dependent variable.

A top down recursive divide and conquer approach sym-

bolizes the tree.Node has two or more branches.Leaf has a
decision or a classification.The topmost node is called the
root node.Decision tree can process both numerical and
categoricaldata. Decision tree is simple to understand and for
visualiza- tion.Not require to prepare data as well.But it is
quite unstable because a small change to data can make a
change to the wholestructure of the tree.

The reached accuracy was,

This curve helps to understand the impact of set of indepen-

dent variables on the dependent variable.This algorithm
5) K-Nearest Neighbour Classifier: This stores all works only when the predicted variable is binary and it
instances in n dimensional space according to training assumes the data is free of missing values and predictors are
data.It works on storing instances of training data and it independent of each other.
doesn’t construct a general internal model.So it is called as a
lazy learningalgorithm. The reached accuracy was,

7) Stochastic Gradient Descendent: It is used to when

there is large number of sample data.It supports different
The classification is determined by a simple majority vote near
number of loss functions and penalties for classification.It
k-nearest neighbors for each point. It is monitored and uses a
calculates derivative from each training data instance and
large number of marked points to mark new points. If a new
calculate the update immediately.
point needs to be labeled, the nearest neighbor (or labeled
point) of the new point is taken into account. The label that
receives the majority of votes from those neighbors will be the
label for the new points. k indicates the number of neighbors
to check. This algorithm is easy to implement.It is robust to
noisy training data.It is efficient to large data sets.But

It is easy to implement.It needs large number of hyper

parameters to increase the efficiency and and the sensitivity
there is no need to compare the value of K and also the cost of the algorithm. There are classification evaluators too.It is the
most impor- tant part to check the efficiency and the accuracy

Volume 5 Issue 11, November 2022 72

www.ijarp.org
 International Journal of Advanced Research and Publications
ISSN: 2456-9992

of the algo- rithm.They are Holdout method, cross validation,

classification report, ROC curve. Cross validation is used in
the research.It supports to decrease over fitting
problem.When K-fold cross validation was conducted
whether the model is over fitted or not can be confirmed.

Data set randomly partitioned into k mutually exclusive

subsets.The sub sets are of same size. One set is kept of
testingand another is kept for training.The process continues
for all the k folds.

Here what has been received for thr model.

Python libraries is used in order to perform classification.

Pandas is the most popular python library for data prepa-
ration and process.It is fast, flexible and easy to use.It
is a data analysis and manipulation tool.It is built on python
programming language. Scikit-learn is also python library
which can be used for machine learning operations.Such as
classification, regression, clustering algorithms.It is a free
software.

E. Identify the complex words from the meaningful

F. Finding related images from web sources for extracted
sentences and substitute it with simple words after that
information
converting that sentence into audio
Finding images from web pages, how to find images from web
In this step identifying complex words from sentences and
pages also done using the same python libraries such as
substituting them with simple words, convert the sentence
newspaper.
into audio was took place.In order to fulfill the requirement
more python package and modules is used.

We used NLTK Wordnet corpus in order to fulfill the

requirment.It is a lexical database of English.

Result of the image scraping is as follow image.

Volume 5 Issue 11, November 2022 73

www.ijarp.org
 International Journal of Advanced Research and Publications
ISSN: 2456-9992

development, clean and pragmatic design.

IV. DISCUSSION
Digital Learning comes with many advantages and dis-
advantages to kids.Advertising is relatively harmful for
kids.Easily kids can be influenced by kids. Because their
critical thinking skills and impulse control aren’t fully de-
veloped.Thus adult contents, food advertising, drugs, cultural
G. Merge audio and images to design and create a biases etc. During Covid-9 most of the educational bodies
digital video production team. faced this type of issues and they had to delay their
teaching.Also with the current economic crisis all the
Deliver digital video production with the combining of transportation has come under limitations and it is same as the
image and audio was the next step.Modern technologies covid pandemic.This kind of system helps in this kind of
is used for this too.Combined images and audio have used to situation for all the educational sector. Parents can turn on
deliver video production. OpenCV is used to develop real- monitor privacy settings on personal devices, apps, social
time computer vision.Also MoviePy i a python modeule media, virtual assistants, and wireless networks.Also they can
which is used for video editing.It can do basic operations like use a plan with children to help guide them toward quality
cut, concatenation, title insertions. Using this real-time media content with fewer ads.Use this plan to talk with
optimized Computer Vision library and MoviePy was children about data collection and media content with fewer
developed a video using the scraped images and generated ads.As well as it is okay to talk to school administrators, and
audio. teachers about avoiding digital products with advertising and
using digital privacy settings on the education technology
H. Mobile application tools they use. As a solution It is been categorized all the
For developing mobile application Flutter was used.It is a keywords into categories which can be easily recognized by
modern mobile app developing tech stack created by the system and stop proceeding for unnecessary contents.
Google.It is open source UI software development kit.It is
used to develop Android, iOS, web apps and many other
from a single codebase. Fig. 1 shows the app UI that was
developed for the digital learning.

I. Web development
React is used mostly in building Front End of web ap-
plications.It is free and open source.It is a javascript
library for building user interfaces based on UI
components.React is maintained by Meta and commiunity
of individual developers and companies.React’s modular
nature makes its code more flexible and easier to maintain
when compared to competing frontend frameworks.
Businesses benefit greatly from this flexibility in terms of
time and money savings. Fig. 2

J. Backend API development

Laravel is used to develop the back end which is also
a PHP Web framework.It is also a free and open source
framework.The majority of web developers are already fa-
miliar with Laravel, the top PHP framework. Most websites
may easily be configured with Laravel’s strong security ca-
pabilities to improve security and defend against hackers
and cybercriminals. For the website, Laravel offers caching
out of the box, which is helpful for speeding up the site.
Laravel offers the ability to create both a straightforward
and expert B2B website and a full-featured eCommerce
website. It may delay completing certain online operations,
including sending emails, until a later time since it employs
a novel message queuing mechanism. Countless third-party
packages are also readily accessible to provide the website a
variety of features and functionality. Django is a high level
web framework based on python and it follows model- Fig. 1. Mobile UI
template-views architecture pattern.It helps to rapid

Volume 5 Issue 11, November 2022 74

www.ijarp.org
 International Journal of Advanced Research and Publications
ISSN: 2456-9992

REFERENCES
[1] “Unesco covid-19 educational disruption and response,”
https://en.unesco.org/covid19/educationresponse.
[2] M. Najar, Alberto, G. Sanzana, A. Grace, Hasan, Amer,
C. Romani,
J. Cristobal, Azevedo, J. P. W. De, Akmal, and Maryam,
“Remote learning during covid-19 : Lessons from today,
principles for tomorrow,” 2022.
[3] Zaharah, Indrayanto, C. Dhien Nourwahidah, A. Saehudin,
H. Hasan, and Kamarusdiana, “The effectiveness of
information technology as a learning media towards
teaching role (case study for student due to pandemic
covid-19),” in 2020 8th International Conference on Cyber
and IT Service Management (CITSM), 2020, pp. 1–6.
[4] A. Tick, “Research on the digital learning and e-learning
behaviour and habits of the early z generation,” in 2018
IEEE 22nd International Con- ference on Intelligent
Engineering Systems (INES), 2018, pp. 000 033– 000 038.
[5] L. Alfaro, C. Rivera, J. Luna-Urquizo, E. Castañeda, J.
Zuniga-Cueva, and M. Rivera-Chavez, “New trends
in e-technologies and e-learning,” in 2021 IEEE World
Conference on Engineering Education (EDUNINE), 2021,
pp. 1–6.
[6] P. Jayawardhana, A. Aponso, N. Krishnarajah, and A.
Rathnayake, “An intelligent approach of text-to-speech
synthesizers for english and sinhala languages,” in 2019
IEEE 2nd International Conference on Information and
Computer Technologies (ICICT), 2019, pp. 229–234.
[7] Z. Yin, “An overview of speech synthesis technology,” in
Fig. 2. Web Application UI 2018 Eighth International Conference on Instrumentation
Measurement, Computer, Communication and Control
V. CONCLUSION (IMCCC), 2018, pp. 522–526.
In this paper a new platform aiming Digital Learning for [8] M.-J. Hwang, R. Yamamoto, E. Song, and J.-M. Kim, “Tts-
kids has been proposed.A lot of research has done on by-tts: Tts- driven data augmentation for fast and high-
how digital learning effect on kids, the technologies can be quality speech synthesis,” in ICASSP 2021 - 2021 IEEE
used for creating the platform and finally the platform is International Conference on Acoustics, Speech and Signal
developed.Artificial Intelligence can be used in the system to Processing (ICASSP), 2021, pp. 6598–6602.
improve more performance.It can be based on the age [9] L. Mathew and V. R. Bindu, “A review of natural language
and the gender.Also Augmented reality can be used to processing techniques for sentiment analysis using pre-
develop the system with more efficiency and user friendly. trained models,” in 2020 Fourth International Conference
NFT or any other tokens can be created and issued to on Computing Methodologies and Communication
encourage the kids as another feature to the proposed system. (ICCMC), 2020, pp. 340–345.
It will add more attraction to the system in technical side. [10] T. P. Nagarhalli, V. Vaze, and N. K. Rana, “Impact of
Almost everyone cannot be interested learn in this machine learning in natural language processing: A
mode.Some part of the crowd may interest in learning phys- review,” in 2021 Third International Conference on
ically, and enjoy meeting others. At the same time who Intelligent Communication Technologies and Virtual
with social anxiety will be interested in learning in online Mobile Networks (ICICV), 2021, pp. 1529–1534.
mode.Also there will be health issues about using digital [11] P. Kłosowski, “Deep learning for natural language
devices for long time such as for eyes and staying in the processing and lan- guage modelling,” in 2018 Signal
same posture for a long time.As kids are always curious in Processing: Algorithms, Architectures, Arrangements, and
learning new things and making observations.So it is Applications (SPA), 2018, pp. 223–228.
required to be extra careful when giving a digital device to [12] L. Safae, B. E. Habib, and T. Abderrahim, “A review of
them.We have to implement all the security measure for machine learning algorithms for web page classification,”
making kids safety. in 2018 IEEE 5th International Congress on Information
Science and Technology (CiSt), 2018, pp. 220– 226.
[13] H. Li, Z. Zhang, and Y. Xu, “Web page classification
method based on semantics and structure,” in 2019 2nd
International Conference on Artificial Intelligence and
Big Data (ICAIBD), 2019, pp. 238–243.

Volume 5 Issue 11, November 2022 75

www.ijarp.org
 International Journal of Advanced Research and Publications
ISSN: 2456-9992

[14] J. Kumar, A. Santhanavijayan, B. Janet, B. Rajendran,

and B. Bindhu- madhava, “Phishing website
classification and detection using machine learning,” in
2020 International Conference on Computer
Communica- tion and Informatics (ICCCI), 2020, pp. Sriram Rajeswaran undergraduate who reading for BSc
1–6. (Hons)in Information Technology at Sri Lankan Institute of
[15] S. Lunn, J. Zhu, and M. Ross, “Utilizing web Information Technology. During 2021-2022 He worked as
scraping and natural language processing to better an UI/UX Intern at a SurgeGlobal. He studied his special
inform pedagogical practice,” in 2020 IEEE Frontiers elective module as Human Computer Interaction and User
in Education Conference (FIE), 2020, pp. 1–9. Experience.
[16] V. Singrodia, A. Mitra, and S. Paul, “A review on
web scrapping and its applications,” in 2019
International Conference on Computer Communication
and Informatics (ICCCI), 2019, pp. 1–6.
[17] D. M. Thomas and S. Mathur, “Data analysis by web
scraping using python,” in 2019 3rd International
conference on Electronics, Commu- nication and
Aerospace Technology (ICECA), 2019, pp. 450–454.

Author Profile

Sharmilan Sureshwaran undergraduate who reading

for BSc (Hons)in Information Technology at Sri Lankan
Institute of Information Technology. During 2021-2022
he worked as an flutter developer Intern at a startup
company called MoodforCode. He studied his special
elective module as Machine Learning

Nilakshana Rajeswaran undergraduate who reading for

BSc (Hons)in Information Technology at Sri Lankan
Institute of Information Technology. During 2021-2022
she worked as an Web developer Intern at a startup
company called MoodforCode. She studied her special
elective module as Machine Learning and Artificial
Intelligence

Kithusshand Raveendran undergraduate who reading

for BSc (Hons)in Information Technology at Sri Lankan
Institute of Information Technology. During 2021-2022
He worked as an Flutter developer Intern at a startup
company called MoodforCode. He studied his special
elective module as Machine Learning.

Volume 5 Issue 11, November 2022 76

www.ijarp.org
View publication stats