Abuse Detection on social media platform twitter using Deep Learning Technique Article 2022

Abuse Detection on social media platform twitter using Deep Learning Technique
(Article 370)
Name of the Student
Name of the University

2
Table of Contents
Abstract............................................................................................................................................3
Literature Review............................................................................................................................4
Abuse on Twitter.............................................................................................................................5
Machine Learning Algorithms:........................................................................................................5
Logistic Regression:....................................................................................................................6
Support Vector Machine:.............................................................................................................6
Random Forest:............................................................................................................................6
XGBoost:.....................................................................................................................................6
Research Questions..........................................................................................................................7
Research Methodology....................................................................................................................7
References........................................................................................................................................8

3
Abstract
Social media has increased connectivity of individuals from all over the world. Social media has
helped in expanding social network of individuals on which they can share their thoughts and
views on several topic. One of these critical topics is Article 370. This topic has been exploded
over various social media platform including Facebook and Twitter. Several people have shared
their thoughts over social media including Twitter. There have been war of words over the
Twitter which include both abusive and non-abusive words. This research has focused on abuse
detection on social media platform utilizing deep learning techniques. Data will be collected
from Twitter using python fetch data techniques. There will be 5 thousand above data based on
high abusive words on Article 370. Classification of 5668 tweets based on abusive and non-
abusive words will be done utilizing sentimental analysis. Several classification models and
machine learning algorithm like Regression, SVM and Random Forest will be utilized in this
research.

4
Literature Review
Jammu and Kashmir is the one and only state in the Indian Union that has been
negotiating its terms of accession. The outcomes of parleys dispersed over months between State
and Central leaderships have been special status provided to state within India which was
legalized under Article 370.1 The Constituent Assembly of India inserted Article 306A on
October 1949 which became operative on November 1952 as Article 370.2 Jammu and Kashmir
was ensured autonomy within Constitution of India. There have been various conflicts happened
based on this Article 370. One of the thoughts, known as the Integrationists, argued this special
position of Jammu and Kashmir and demanded full integration of the State in the India. Its
slogan started with ‘Ek Pradhan, Ek Vidhan, Ek Nishan’ (One PM, One Constitution, One Flag).
Article 370 come up with great power to BJP leader and other anti-autonomy groups as it
provide documentary evidence for showing Article 370 had a full approval of the late Shyama
Prasad Mukherjee.3 This article states that residents of Jammu and Kashmir live under a separate
set of laws including citizenship, ownership of property and fundamental rights. President Ram
Nath Kovind issued a constitutional order suspending the 1954 order and making all provisions
of the Indian constitution applicable to Jammu and Kashmir on 5 August 2019.4
The imposition of Article 370 have created a huge blast of comments and sharing over social
media platforms. Various politician and civilians have been sharing their views and thoughts
1 Bhat, K.A., (2017). Special Status of Jammu & Kashmir: Article 370: An Indepth Analysis. Educreation Publishing.
2 Farooqi, S., & Shafiq, Z. (2019). Measurement and Early Detection of Third-Party Application Abuse on Twitter.
3 Saxena, R., (2018). Asymmetrical Federalism in India: Promoting Secession or Accommodating Diversity?. In Revisiting Unity
& Diversity in Federal Countries (pp. 362-376). Brill Nijhoff.
4 Gupta, A. & Kaushal, R., (2015), March. Improving spam detection in online social networks. In (2015) International
conference on cognitive computing & information processing (CCIP) (pp. 1-6). IEEE.

5
over the social media. Social Media has helped in spreading news and information regarding
Article 370. Reactions and Views of people and politicians have been channelized by the use of
Social Media including Twitter.
Abuse on Twitter
Reactions of people and politicians have been high over the Twitter. It is a globalized
social media platform that is utilized by most of people and politicians. Individuals can rate their
comment about several products and services on various backgrounds. India has the largest
democracy in all over the world.5 Therefore, it becomes overrated during elections and Twitter
become the weapon for political leaders and party to promote their opinion.6 This become
tremendous abusive over the Twitter. In this case of Article 370, Twitter has been overloaded
with millions of tweets and comments of people and political leaders. People used to spread
hatred and abusive words over the Twitter.7 There can be classification of abusive and non-
abusive words over Twitter. This can be done using Sentimental Analysis.
Machine Learning Algorithms:
The four machine learning algorithm discussed in the research are as follows:
5 Hassanpour, S., Tomita, N., DeLise, T., Crosier, B. & Marsch, L.A., (2019). Identifying substance use risk based on deep
neural networks & Instagram social media data. Neuropsychopharmacology, 44(3), p.487.
6 Chen, H., McKeever, S. & Delany, S.J., (2019), July. The Use of Deep Learning Distributed Representations in the
Identification of Abusive Text. In Proceedings of the International AAAI Conference on Web & Social Media (Vol. 13, No. 01,
pp. 125-133).
7 Aroyehun, S.T. & Gelbukh, A., (2018), August. Aggression detection in social media: Using deep neural networks, data
augmentation, & pseudo labeling. In Proceedings of the First Workshop on Trolling, Aggression & Cyberbullying (TRAC-
(2018)) (pp. 90-97).

6
Logistic Regression:
Logistic regression is termed to be as most genuine and consistence regression analysis in
classification problem which conducts when the dependent variables are dichotomous. It is also
called sigmoid function. Logistic regression is an estimating parameters of a logistic model in
logistic regression. The outcome of the logistic regression is measured with a dichotomous
variable.
Support Vector Machine:
Support vector machine are very specific class of regression algorithms. The support vector
machine can be applied not only to the classification problem but can also be applied to
regression problems. The hyper plane in the 2-D space is basically a line dividing a particular
plane into two distinct and unique parts where in each class lay in either side.
Random Forest:
Random forest is termed to be one of the most effective machine learning model for predictive
analytics. Random forests is a crucial ensemble learning method for the classification then
regression and other tasks. Ensemble methods are supervised learning model which combines
different prediction to improve the prediction accuracy and generalization of multiple smaller
model. Random model is also a type of additive model.
XGBoost:
XGBoost is best known to provide better solutions to any models other than any of the machine
learning algorithms. XGBoost is a special implementation of gradient boosted decision tree
designed to provide the speed and performance. It implements machine learning algorithms
under the Gradient Boosting framework.

7
Research Questions
Following are the research questions:
 How can abusive and non-abusive words be detected utilizing classification models of
deep learning from Twitter comments?
 Which classification model of deep learning can be utilized for detecting abusive and
non-abusive words from Twitter comments?
Research Methodology
Data will be collected from Twitter API over which 5 thousand data will be collected
from twitter on Article 370 comments on India. Data will be collected in form of hashtags, URL,
@ symbol. Later on, only English text will be classified and codes will be removed from any
other languages. Only abusive comments will be classified from overall data collected from
Twitter. Data will be analysed using sentimental analysis method.
The research will follow all ethical norms under the academic research guidelines. Data
and Information will be collected after approval granted from Twitter. Details of person who
have commented will not be disclosed in the research. There will be no tampering of data and
information one during analysis of data. During sentiment analysis, no outcomes will focus on
hurting someone’s feelings and thought about the research topic. Results and outcomes of
sentiment analysis will be focused on only research purpose. Security and safety of data will be
done under the Data Protection Act 1998.

8
References
Bhat, K.A., (2017). Special Status of Jammu & Kashmir: Article 370: An Indepth Analysis.
Educreation Publishing.
Saxena, R., (2018). Asymmetrical Federalism in India: Promoting Secession or Accommodating
Diversity?. In Revisiting Unity & Diversity in Federal Countries (pp. 362-376). Brill Nijhoff.
Hassanpour, S., Tomita, N., DeLise, T., Crosier, B. & Marsch, L.A., (2019). Identifying
substance use risk based on deep neural networks & Instagram social media
data. Neuropsychopharmacology, 44(3), p.487.
Aroyehun, S.T. & Gelbukh, A., (2018), August. Aggression detection in social media: Using
deep neural networks, data augmentation, & pseudo labeling. In Proceedings of the First
Workshop on Trolling, Aggression & Cyberbullying (TRAC-(2018)) (pp. 90-97).
Founta, A.M., Chatzakou, D., Kourtellis, N., Blackburn, J., Vakali, A. & Leontiadis, I., (2019),
June. A unified deep learning architecture for abuse detection. In Proceedings of the 10th ACM
Conference on Web Science (pp. 105-114). ACM.
Wu, T., Liu, S., Zhang, J. & Xiang, Y., (2017), January. Twitter spam detection based on deep
learning. In Proceedings of the australasian computer science week multiconference (p. 3).
ACM.
Chen, H., McKeever, S. & Delany, S.J., (2019), July. The Use of Deep Learning Distributed
Representations in the Identification of Abusive Text. In Proceedings of the International AAAI
Conference on Web & Social Media (Vol. 13, No. 01, pp. 125-133).
Gupta, A. & Kaushal, R., (2015), March. Improving spam detection in online social networks.
In (2015) International conference on cognitive computing & information processing (CCIP) (pp.
1-6). IEEE.

9
Farooqi, S., & Shafiq, Z. (2019). Measurement and Early Detection of Third-Party Application
Abuse on Twitter.
Singh, V., Varshney, A., Akhtar, S., Vijay, D., & Shrivastava, M. (2018). of the Second
Workshop on Abusive Language Online (ALW2), pages 43–50 Brussels, Belgium, October 31,
2018. c 2018 Association for Computational Linguistics 43 Aggression Detection on Social
Media Text Using Deep Neural Networks.