Comparing Machine Learning and Deep Learning Techniques for Text Analytics - Detecting the Severity of Hate Comments Online

Aug 5, 2026

Social media platforms enable users to share views publicly, yet abusive and harassing comments increasingly undermine individuals’ wellbeing. This study introduces a novel approach to combat online harassment by detecting the severity of abusive comments on social platforms. It compares machine learning models (Naïve Bayes, Random Forest, Support Vector Machine) with deep learning models (CNN and Bi-LSTM), while examining the impact of text pre-processing. Features are built using unigrams and bigrams for machine learning and word embeddings for deep learning. Random Forest with bigrams achieves the best overall performance, supporting an efficient severity detection model for hate language online with theoretical and practical implications.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages19

Reading Time48 min

AccessFree to download

Introduction

Social media and online harassment

Importance of detecting online abuse

Dark sides of social media and hate speech research

Related Work

Prior approaches and limitations in harassment detection

Methodology

Compared machine learning and deep learning models

Text pre-processing and feature representation

Results

Performance comparison and best model selection

Discussion

Implications for theory and practice

Conclusion

Contributions and future research directions

Information Systems Frontiers [https://doi.org/10.1007/s10796-023-10446-x](https://doi.org/10.1007/s10796-023-10446-x) Comparing Machine Learning and Deep Learning Techniques for Text Analytics: Detecting the Severity of Hate Comments Online Alaa Marshan1 · Farah Nasreen Mohamed Nizar2 · Athina Ioannou3 · Konstantina Spanaki4 Accepted: 1 November 2023 © The Author(s) 2023 Abstract Social media platforms have become an increasingly popular tool for individuals to share their thoughts and opinions with other people. However, very often people tend to misuse social media posting abusive comments. Abusive and harassing behaviours can have adverse effects on people's lives. This study takes a novel approach to combat harassment in online platforms by detecting the severity of abusive comments, that has not been investigated before. The study compares the performance of machine learning models such as Naïve Bayes, Random Forest, and Support Vector Machine, with deep learning models such as Convolutional Neural Network (CNN) and Bi-directional Long Short-Term Memory (Bi-LSTM) . Moreover, in this work we investigate the effect of text pre-processing on the performance of the machine and deep learning models, the feature set for the abusive comments was made using unigrams and bigrams for the machine learning models and word embeddings for the deep learning models. The comparison of the models’ performances showed that the Random Forest with bigrams achieved the best overall performance with an accuracy of (0.94), a precision of (0.91), a recall of (0.94), and an F1 score of (0.92) . The study develops an efficient model to detect severity of abusive language in online platforms, offering important implications both to theory and practice. Keywords Machine learning · Deep learning · Hate speech · Social media · Text pre-processing · Text representation · Text analytics 1 Introduction The introduction of social media has significantly affected people’s lives, allowing them to publicly share their opinions and beliefs about various issues spanning areas such as politics, economics, health and social issues (Meske & Bunde, 2022) . As a result, social media has been deemed as one of the most prominent contributors to the freedom of speech principle (Putri et al., 2020). However, recently freedom of speech has been abused on many occasions, bringing negative consequences both to the individuals themselves * Alaa Marshan[a.marshan@surrey.ac.uk](a.marshan@surrey.ac.uk) 1 Department of Computer Science, University of Surrey, Guildford, UK 2 Department of Computer Science, Brunel University, London, UK 3 Surrey Business School, University of Surrey, Guildford, UK 4 Audencia Business School, Nantes, France as well as others who are being abused (Putri et al., 2020) . Formally defined as online harassment, online abuse refers to verbal and/or graphical abuse towards others on an online platform (Karatsalos & Panagiotakis, 2020) . Online harassment can adversely affect people's lives as people being targeted by others feel physical and mental suffering (Modha et al., 2020a) . According to Matamoros-Fernández and Farkas,(2021, p. 205),‘sociality is continuously transformed by the interplay of humans and technology’. As such, social media platforms, although characterised as communication infrastructures that are quite open and decentralised, where users can widely share their opinions, participate, and develop new networks (e.g. , activism); they have also amplified several forms of abuse, such as digital hate speech, racism, and online discrimination (Kim et al. , 2022 ; Matamoros-Fernández & Farkas, 2021) . The spread of digital hate speech through social media has evidently contributed to the reshaping of“racist dynamics through their affordances, policies, algorithms and corporate decisions”(Matamoros-Fernández & Farkas, 2021, 206) . Evidence from previous studies demonstrates the widespread reshaping of struct

FAQs

What problem does the study address on social media?

The study targets misuse of social media through abusive and harassing comments, aiming to detect how severe such comments are.

Which models are compared for detecting severity of hate comments?

It compares machine learning models—Naïve Bayes, Random Forest, and Support Vector Machine—with deep learning models including a Convolutional Neural Network (CNN) and Bi-directional Long Short-Term Memory (Bi-LSTM).

How do text pre-processing and features affect model performance?

The study evaluates the effect of text pre-processing and uses unigrams and bigrams for machine learning, while deep learning relies on word embeddings for text representation.

Aug 5, 2026

Upload by Olivia Brown