DESIGN AND IMPLEMENTATION OF AUTOMATED SOCIAL MEDIA HATE SPEECH DETECTION SYSTEM

105

ABSTRACT

Social Media Hate Speech has continued to grow both locally and globally due to the increase of Online Social Media web forums like Facebook, Twitter and blogging. This has been propelled even further by smartphones and mobile data penetration locally. Global and Local terrorism has posed a vital question for technologists to investigate, prosecute, predict and prevent Social Media Hate Speech.

This study provides a social media digital forensics tool through the design, development and implementation of a software application. The objective of this study is to identify and analyze the different software techniques for hate speech monitoring and provide the best suited and customized application. The study will develop an application using Linux Apache MySQL PHP and Python. The application will use Scrapy Python page ranking algorithm to perform web crawling and the data will be placed in a MySQL database for data mining.

The application used Agile Software development methodology with twenty websites being the subject of interest. The websites will be the sample size to demonstrate how the application works together with the Python libraries as the framework for web crawling. MySQL data mining, database query application models will be used in performing the search of the lexicon of keywords for hate speech, Inferences from the data mined from crawled web pages will be drawn using Microsoft Excel 2016. Excel will be used for data analysis with the data being presented in tables and figures.

AUTOMATED  SOCIAL MEDIA HATE SPEECH DETECTION USING DATA MINING TECHNIQUE

CHAPTER ONE/INTRODUCTION

The Internet is continually providing the medium for communication, fundraising, training, media operations and recruitment. Web forums are online Internet websites where the owners of the webpage can post stories, ideologies, blogs; which in turn undergoes discussions from the community members and participants. Web forums such as Skype, Gtalk, WhatsApp, Twitter and Facebook have become valuable arenas for social interaction, discussion and communication online on the Internet.

Internet accessibility in Kenyan has continued to penetrate and grow in numbers, with 52% of the total population of 42 Million citizens being connected. In comparison to other countries in the East Africa bloc, Kenya’s Internet usage and penetration is higher and this has mainly been facilitated by the introduction of smartphones and cheap mobile data bundles provided by Internet Service Providers and mobile networks. Currently, Kenya’s mobile tele density is placed at 77% facilitated by three mobile service providers, that is, Safaricom, Airtel and Orange (Nanfuka, 2014).

Obar (2015) defined social media as computer-mediated tools for creating, sharing and exchanging of information, photos, videos and career interests in the virtual communities and networks. Social Media services have been defined as Web 2.0 applications and web forums that emphasize on online generated content and interoperability (Darcy, 1999). This has led to web forums, blogs, wiki and media-sharing websites that grant online users the capability to communicate and respond to opinions on daily news, political debates, tribal and religion discussions.

Sutton (2009) formulates social media as diverse and developing online communication that allows the production and sharing of information (Sutton, 2009). Crisis communication, public relations, radicalization and hate speech has been acknowledged as the roles by which social media is aiding online communication. Social media forums such as Twitter have played a daily constant role in issuing daily political news, traffic updates, emergency warnings and alerts, supporting recovering efforts, situational awareness and real-time information.

Social Media, web forums and online conversations and debates have brought to the fore hate speech. The hate speech is due to the ethnic tensions which have polarized the country with each passing political season. Hate speech was in the limelight during the 2007 post-election results, as more citizens were using phone short messages to fuel, share destructive propaganda and support ethnic based violence of the different tribal communities depending on where you were aligned. ecame most apparent during the 2007 post-election period during which short message service (SMS) fueled political conflict and ethnic based violence. The number of reported causalities were not more than 1,200 and aleast 500,000 people being internally displaced. Current studies have shown the movement of hate speech from traditional person to person phone message in to social media platforms such as Facebook and Twitter. There was a swift change of political bickering and political tension from the physical to the digital cyber space. This was fueled much by the almost instant delivery of propaganda, agenda and tribal ethnic through social media forums such as Twitter, WhatspApp and Facebook.

The 2013 presidential debates and elections saw the maximization of social media for campaign purposes, voter communication and civic engagement. This in turn saw the resurgence of online hate speech.

Operation Linda Nchi, a codename for the joint coordinated military operation of the Kenyan and Somali armies began on 16 October 2011 (Kron, 2013) in Somalia. This is after it had been believed that Al-Shaabab militants had kidnapped several foreign aid workers and tourists visiting the country in Lamu, a coastal town. Web forums have been used by radical groups for communication, sharing of information and their ideologies to the public. Such web forums form part of the Dark Web.

The Dark Web is made up of web forums, websites, chat groups and social media platforms that are used by terrorists, hate mongers, radicals, political, ethnic and extremist groups (Chen, 2006). The war on terrorism has been both a physical, psychological and virtual battle, with extremists using web forums to spread and influence their propaganda.

Hate speech is characterized by any form of verbal or non-verbal attack targeting a specific group of people. It can be motivated by racial, ethnic, gender, religion or sexual orientation. Hate speech is usually communicated through different media such as Internet, hand-held devices, newspapers,magazines, television, radio broadcasts, verbal person-to-person.

Hate Speech is defined by the National Cohesion and Integration 2008 Act as “speech that is threatening, abusive or insulting or involves the use of threatening, abusive or insulting words” with the intention to stir up ethnic hatred or a likelihood that ethnic hatred will be stirred up. Hate Speech can be ethical or also radical when used by extremist on the basis of religion for recruiting people in terrorist acts. Hate speech is continuously evolving to the digital space and its being embodied on the social media through ethnical and radical groups and classification.

Social Media Forensics for Hate Speech Opinion Mining relates to the scientific application of cyber forensics tools to social media web forums in order to extract, identify and document hate speech.

Hate Speech on Social Media fueled the 2007 post-election violence witnessed in the country. It was one of the known contributors for ethnic strife and tension. The Law Enforcement and Communication authorities need to setup the war on the vice so as to avoid future problems. Social Media and Crime has shown a clear relationship that someone knew of the possible crime or planning of the crime. Wilson et. al (2012) argued that “Facebook and other social media platforms need to be positioned firmly within the criminologist’s gaze not only because of the wealth of data these platforms provide but also due to their significant popularity across all age groups and their influence on how people interact and communicate with one another. This is because social media is now intimately interacting with many of the issues that we as criminologists can and should be engaging with”.

The amount of big data that social media generates is immense, instant, dynamic and constantly evolving. The current law enforcement agencies and communication authorities are stretched and the need to provide an automated system is paramount.

The movement of extremist group towards the use of social media for recruiting Al-Shaabab militants is also of grave concern, as the recruiters are aware that there are no checks on the social media and cyber space. Currently, the security agencies depend on public to notify them of any hate speech comments on the Internet. The cyber hate mongers are quick enough to remove and de-register their social media accounts. Hence, the need of a solution which is more proactive and keeps a copy of the collected digital evidence.

 Problem Statement

Hate Speech on social media needs to be tracked, tackled and means provided to apprehend the cyber criminals and hate speech mongers. The process of collecting and documenting online hate speech digital evidence should be optimized efficiently.

With security agencies and communication authority personnel being overstretched, with less manpower and tools, cyber expertise, the need for an automated and easier to use system for hate speech is crucial.

Using the Social Media platform, we will provide an application that will be able to mine social media opinions and easily present the results of possible hate speech crime together with the evidence and potential list of suspects. The system will invoke a web crawler that will collect all the web forum details and insert them in a database. Once the database is created, a script will be created to check the data against a list of potential keywords of hate speech. The script will search and provide the list of hate speech data together with the person of interest.

Communication authorities, security agencies and the country at large stands to benefit by placing checks on social media hate speech. The movement of hate speech mongers towards the digital cyber space needs to be addressed before it escalates further as was experienced regularly when political debates are held or government corruption cases are discussed.

Objectives of the Study 

The purpose of the study was to design, develop and implement a software application for hate speech monitoring and reporting.

 Specific Objectives

  1. To identify and analyze techniques used in hate speech monitoring and select the best suitable technique for creating a customized hate speech
  2. To develop an application that will combine hate speech keywords for data
  3. To demonstrate and test the application while providing analysis on the hate speech websites being

  Research Questions

The research answers the following:

  1. How can Social Media web forums be monitored and forensic digital evidence collected?
  2. What are the digital investigative challenges that law enforcement are facing towards evidence carving?
  3. What are the methods used to collect the hate speech opinions and developing an algorithm?
______________________________________