Modeling Hate Speech Detection in Social Media Interactions Using Bert

International Journal of Research and Scientific Innovation (IJRSI) | Volume VII, Issue II, February 2020 | ISSN 2321–2705

Modeling Hate Speech Detection in Social Media Interactions Using Bert

Gibran Mwadime¹, Moses Odeo², Boniface Ngari³, Stephen Mutuvi^4
1,2,3,4Department of Computer Science, Multimedia University of Kenya

Abstract—Hate speech propagation in social media sites has been happening over time and there is need to accurately identify and counter it so that those offended can seek redress and offenders can be punished for perpetrating the vice. In this paper, we demonstrate how fine tuning a pre-trained Google Bidirectional Encoder Representation from Transformers (BERT) model has been used to achieve an improvement in accuracy of classification of tweets as either hate speech or not. Random forests and logistic regression algorithms have been used to build baseline models with a publicly available twitter dataset from hatebase.org. To validate the BERT model, we collected data using Tweepy API and combined with data from hatebase.org for training. The results obtained show an improvement in accuracy of tweets classification as either hate speech or not from the baseline models by 7.22%.

Keywords: Sentiment analysis, hate speech, social media, model, data

I. INTRODUCTION

Hate speech is defined as any form of expression that seeks or expresses hatred against a person or group of persons because of something they are associated with [6]. Social media has become an environment where people from all walks of life and from different geographical locations converge to share experiences, opinions and ideas. In the recent years, there has been a huge growth in the use of social media and misuse to propagate hate speech and related activities [7]. A lot of information is generated that contains ambiguities and a lot of noise making it difficult to decipher what it means. This calls for a reliable and accurate sentiment analysis tool to ensure text is understood clearly and categorized to the class it belongs appropriately. Differences in opinions have led to abuse and exchanges that result in hatred between parties involved. In the UK, moments after Butt used a van to run over pedestrians, twitter was abuzz with over 18 million tweets published within an hour containing happy messages, hatred towards Islam and support for violence [2].

Read More