Automatic Hate Speech Detection Research Articles

Roman urdu hate speech detection using hybrid machine learning models and hyperparameter optimization.

With the rapid increase of users over social media, cyberbullying, and hate speech problems have arisen over the past years. Automatic hate speech detection (HSD) from text is an emerging research problem in natural language processing (NLP). Researchers developed various approaches to solve the automatic hate speech detection problem using different corpora in various languages, however, research on the Urdu language is rather scarce. This study aims to address the HSD task on Twitter using Roman Urdu text. The contribution of this research is the development of a hybrid model for Roman Urdu HSD, which has not been previously explored. The novel hybrid model integrates deep learning (DL) and transformer models for automatic feature extraction, combined with machine learning algorithms (MLAs) for classification. To further enhance model performance, we employ several hyperparameter optimization (HPO) techniques, including Grid Search (GS), Randomized Search (RS), and Bayesian Optimization with Gaussian Processes (BOGP). Evaluation is carried out on two publicly available benchmarks Roman Urdu corpora comprising HS-RU-20 corpus and RUHSOLD hate speech corpus. Results demonstrate that the Multilingual BERT (MBERT) feature learner, paired with a Support Vector Machine (SVM) classifier and optimized using RS, achieves state-of-the-art performance. On the HS-RU-20 corpus, this model attained an accuracy of 0.93 and an F1 score of 0.95 for the Neutral-Hostile classification task, and an accuracy of 0.89 with an F1 score of 0.88 for the Hate Speech-Offensive task. On the RUHSOLD corpus, the same model achieved an accuracy of 0.95 and an F1 score of 0.94 for the Coarse-grained task, alongside an accuracy of 0.87 and an F1 score of 0.84 for the Fine-grained task. These results demonstrate the effectiveness of our hybrid approach for Roman Urdu hate speech detection.

Automatic hate speech detection in audio using machine learning algorithms

Automatic hate speech detection in audio using machine learning algorithms

Automatic Hate Speech Detection and the hassle of Offensive Language

A key task for automatic hate-speech detection on social media is the separation of hate speech from different instances of offensive language. Lexical detection strategies tend to have low precision due to the fact they classify all messages containing precise terms as hate speech and previous work the use of supervised gaining knowledge of has failed to differentiate among the two classes. We used a crowd-sourced hate speech lexicon to acquire tweets containing hate speech keywords. We use crowdsourcing to label a pattern of those tweets into three classes: those containing hate speech, only offensive language, and those with neither. We educate a multi-magnificence classifier to distinguish among those one-of-a-kind categories. near analysis of the predictions and the errors suggests when we can reliably separate hate speech from different offensive language and while this differentiation is extra difficult. we discover that racist and homophobic tweets are much more likely to be categorized as hate speech but that sexist tweets are normally labeled as offensive. Tweets without specific hate key phrases also are more difficult to categories.

Hate Speech Detection in Arabic Text: Survey

In light of the continuous expansion of social media content, users have the freedom to express themselves without any boundaries or restrictions over the content. Unfortunately, this has increased the spread of hateful speech among users, which in turn has led to an increase in crimes, murders, and even acts of terrorism. Differentiating hate speech from other offensive language poses a significant challenge, requiring deep linguistic analysis. The task is further complicated by the scarcity of Arabic data resources, which significantly limits the availability of relevant information and the Arabic morphology's richness. This study seeks to conduct a comprehensive review of the current state of research on hate speech-related issues and automatic hate speech detection in Arabic text on social media platforms.

T5 for Hate Speech, Augmented Data, and Ensemble

We conduct relatively extensive investigations of automatic hate speech (HS) detection using different State-of-The-Art (SoTA) baselines across 11 subtasks spanning six different datasets. Our motivation is to determine which of the recent SoTA models is best for automatic hate speech detection and what advantage methods, such as data augmentation and ensemble, may have on the best model, if any. We carry out six cross-task investigations. We achieve new SoTA results on two subtasks—macro F1 scores of 91.73% and 53.21% for subtasks A and B of the HASOC 2020 dataset, surpassing previous SoTA scores of 51.52% and 26.52%, respectively. We achieve near-SoTA results on two others—macro F1 scores of 81.66% for subtask A of the OLID 2019 and 82.54% for subtask A of the HASOC 2021, in comparison to SoTA results of 82.9% and 83.05%, respectively. We perform error analysis and use two eXplainable Artificial Intelligence (XAI) algorithms (Integrated Gradient (IG) and SHapley Additive exPlanations (SHAP)) to reveal how two of the models (Bi-Directional Long Short-Term Memory Network (Bi-LSTM) and Text-to-Text-Transfer Transformer (T5)) make the predictions they do by using examples. Other contributions of this work are: (1) the introduction of a simple, novel mechanism for correcting Out-of-Class (OoC) predictions in T5, (2) a detailed description of the data augmentation methods, and (3) the revelation of the poor data annotations in the HASOC 2021 dataset by using several examples and XAI (buttressing the need for better quality control). We publicly release our model checkpoints and codes to foster transparency.

An approach to automatic classification of hate speech in sports domain on social media

Hate Speech encompasses different forms of trolling, bullying, harassment, and threats directed against specific individuals or groups. This phenomena is mainly expressed on Social Networks. For sports players, Social Media is a means of communication with the widest part of their fans and a way to face different cyber-aggression forms. These virtual attacks can harm players, distress them, cause them to feel bad for a long time, or even escalate into physical violence. To date, athletes were not observed as a vulnerable group, so they were not a subject of automatic Hate Speech detection and recognition from content published on Social Media. This paper explores whether a model trained on the dataset from one Social Media and not related to any specific domain can be efficient for the Hate Speech binary classification of test sets regarding the sports domain. The experiments deal with Hate Speech detection in Serbian. BiLSTM deep neural network was learned with different parameters, and the results showed high Precision of detecting Hate Speech in sports domain (96% and 97%) and pretty low Recall.

Accelerating automatic hate speech detection using parallelized ensemble learning models

With increasing number of social media users and online engagement, it is essential to study hate speech propagation on social media platforms (SMPs). Automatic hate speech detection on social media is of utmost importance as hate speech can create discomfort among users and potentially generate a strong reaction in society. Ensemble learning algorithms are helpful in addressing sentiment-based classification due to their fault tolerance and efficiency. However, a simple, scalable, and robust framework is required to deal with large-scale data efficiently and accurately. Therefore, we propose parallelization to the standard ensemble learning algorithms to speed up the automatic hate speech detection on SMPs. In this study, we parallelize bagging, A-stacking, and random sub-space algorithms and test their serial and parallel versions on the standard high-dimensional datasets for hate speech detection. The experiments are performed using six datasets that address hate speech propagation during events like the COVID-19 pandemic, the US presidential election (2020), and the farmers’ protest in India (2021). Our parallel models observe a significant speedup with high efficiency, claiming that the proposed models are suitable for the considered application. Also, one of the main motivations of this study is to highlight the importance of generalization by testing the models under the cross-dataset environment. We observed that the accuracy is not affected while parallelizing the algorithms compared with serial algorithms executing on a single machine.

Exploring Automatic Hate Speech Detection on Social Media: A Focus on Content-Based Analysis

Hate speech is a challenging problem, and its dissemination can cause potential harm to individuals and society by creating a sense of general unwelcoming to the marginalized groups, which usually are targeted. Therefore, it is essential to understand this issue and which techniques are useful for automatic detection. This paper presents a survey on automatic hate speech detection on social media, providing a structured overview of theoretical aspects and practical resources. Thus, we review different definitions of the term “hate speech” from social network platforms and the scientific community. We also present an overview of the methodologies used for hate speech detection, and we describe the main approaches currently explored in this context, including popular features, datasets, and algorithms. Furthermore, we discuss some challenges and opportunities for better solving this issue.

Automated offensive language classification through “emotional” comments from network users

In recent years, online hate speech has become widespread as a result of the ease with which any online opinion can be presented on social media. This trend is mainly due to the rapid growth of mobile devices, computers and the Internet. There is a lot of interest in research on automatic hate speech detection. However, there is not a lot research on whether social networking sites can help identify hate crimes and related communities. Hate speech has a negative impact on the mental level of any person. However, some people are more prone to emotional instability than others. It is difficult for racial or ethnic groups whose language is not popular automatically collecting and evaluate texts created on the Internet. In this article, using artificial intelligence techniques in text processing and natural language processing, for the first time, a database was collected using various keywords to identify offensive words in social networks.

Assessing the Impact of Contextual Information in Hate Speech Detection

In recent years, hate speech has gained relevance in social networks and other virtual media because of its intensity and its relationship with violent acts against members of protected groups. Due to the huge amounts of user-generated contents, a great effort has been made to develop automatic tools to aid the analysis and moderation of this speech, at least in its most threatening forms. One of the limitations of current approaches to automatic hate speech detection is the lack of context; most studies and resources focus on isolated messages, without considering any type of conversational context or even the topic being discussed. This severely restricts the available information for determining whether a post on a social network should be tagged as hateful or not. In this work, we assess the impact of adding contextual information to the hate speech detection task. In particular, we study a Twitter subdomain consisting of replies to posts by news outlets, which provides a natural environment for contextualized hate speech detection.We collected a novel corpus in the Rioplatense dialectal variety of Spanish focusing on hate speech associated with the COVID-19 pandemic, and manually annotated it using carefully designed guidelines. Our classification experiments using state-of-the-art transformer-based machine learning techniques show evidence that adding contextual information improves the performance of hate speech detection for two proposed tasks (binary and multi-label prediction), increasing their Macro F1 by 4.2 and 5.5 points, respectively. These results highlight the importance of exploiting contextual information for the task of hate speech detection. We make our code, models, and corpus available for further research.

Automatic hate speech detection using aspect based feature extraction and Bi-LSTM model

Automatic hate speech detection using aspect based feature extraction and Bi-LSTM model

Quantifying the impact of context on the quality of manual hate speech annotation

AbstractThe quality of annotations in manually annotated hate speech datasets is crucial for automatic hate speech detection. This contribution focuses on the positive effects of manually annotating online comments for hate speech within the context in which the comments occur. We quantify the impact of context availability by meticulously designing an experiment: Two annotation rounds are performed, one in-context and one out-of-context, on the same English YouTube data (more than 10,000 comments), by using the same annotation schema and platform, the same highly trained annotators, and quantifying annotation quality through inter-annotator agreement. Our results show that the presence of context has a significant positive impact on the quality of the manual annotations. This positive impact is more noticeable among replies than among comments, although the former is harder to consistently annotate overall. Previous research reporting that out-of-context annotations favour assigning non-hate-speech labels is also corroborated, showing further that this tendency is especially present among comments inciting violence, a highly relevant category for hate speech research and society overall. We believe that this work will improve future annotation campaigns even beyond hate speech and motivate further research on the highly relevant questions of data annotation methodology in natural language processing, especially in the light of the current expansion of its scope of application.

Improving Zero-Shot Cross-Lingual Hate Speech Detection with Pseudo-Label Fine-Tuning of Transformer Language Models

Hate speech has proliferated on social media platforms in recent years. While this has been the focus of many studies, most works have exclusively focused on a single language, generally English. Low-resourced languages have been neglected due to the dearth of labeled resources. These languages, however, represent an important portion of the data due to the multilingual nature of social media. This work presents a novel zero-shot, cross-lingual transfer learning pipeline based on pseudo-label fine-tuning of Transformer Language Models for automatic hate speech detection. We employ our pipeline on benchmark datasets covering English (source) and 6 different non-English (target) languages written in 3 different scripts. Our pipeline achieves an average improvement of 7.6% (in terms of macro-F1) over previous zero-shot, cross-lingual models. This demonstrates the feasibility of high accuracy automatic hate speech detection for low-resource languages. We release our code and models at https://github.com/harisbinzia/ZeroshotCrosslingualHateSpeech.

Towards a Benchmarking System for Comparing Automatic Hate Speech Detection with an Intelligent Baseline Proposal

Hate Speech is a frequent problem occurring among Internet users. Recent regulations are being discussed by U.K. representatives (“Online Safety Bill”) and by the European Commission, which plans on introducing Hate Speech as an “EU crime”. The recent legislation having passed in order to combat this kind of speech places the burden of identification on the hosting websites and often within a tight time frame (24 h in France and Germany). These constraints make automatic Hate Speech detection a very important topic for major social media platforms. However, recent literature on Hate Speech detection lacks a benchmarking system that can evaluate how different approaches compare against each other regarding the prediction made concerning different types of text (short snippets such as those present on Twitter, as well as lengthier fragments). This paper intended to deal with this issue and to take a step forward towards the standardization of testing for this type of natural language processing (NLP) application. Furthermore, this paper explored different transformer and LSTM-based models in order to evaluate the performance of multi-task and transfer learning models used for Hate Speech detection. Some of the results obtained in this paper surpassed the existing ones. The paper concluded that transformer-based models have the best performance on all studied Datasets.

Hate Speech Detection Using Text Mining and Machine Learning

Automatic hate speech detection on social media is becoming an outstanding concern in modern countries. Indeed, hate speech towards people brings about violent acts and social chaos, hence law prohibits it, and it engenders moral and legal implications. It is crucial that we can precisely categorize the hate speech, and not a hate speech automatically, while this allows us to identify easily real people who represent a threat for our society, and who wrongly regard as hateful speakers. In this paper, we applied a complete text mining process and Naïve Bayes machine learning classification algorithm to two different data sets (tweets_Num1 and tweets_Num2) taken from Twitter, to better classify tweets. The results obtained demonstrate that our model performed well regarding different metrics based on the confusion matrix including the accuracy metric, which achieved 87. 23% on the first dataset, and 93. 06% on the second.

Large Comparative Study of Recent Computational Approach in Automatic Hate Speech Detection

Social media has become a constant in our everyday life. However, its steady growth has increased the hate speech and hostile content problem. To curb this, hate speech detection and recognition is required, but it is faced to two major challenges - laws and enforcement, and automatic computerized hate speech detection. Although many studies are already implemented in detecting hate content, many of these are done in a single setting showing a single dataset in comparison to machine learning or deep learning models. Thus, there is no comparison between previous approaches and recent inventions such as transformer model. Therefore, in this work we explored and compared recent advanced approaches in automatic hate speech detection. Our aim is to analyze the influence different approaches in detecting hate content and its applicability in the real world. Several experiments were conducted on eight real hate speech datasets from recent studies. We present the results of each comparison which shows that the recent transformer model approach is able to outmatch many of the previous hate speech recognition models by significant G-Means and F1 scores. To the author’s knowledge, this paper is the first attempt to present a large comparative study of approaches in hate speech detection.

Detecting Hate Speech on Twitter Network using Ensemble Machine Learning

Twitter is habitually exploited now-a-days to propagate torrents of hate speeches, misogynistic, and misandry tweets that are written in slang. Machine learning methods have been explored in manifold studies to address the inherent challenges of hate speech detection in online spaces. Nevertheless, language has subtleties that can make it stiff for machines to adequately comprehend and disambiguate the semantics of words that are heavily dependent on the usage context. Deep learning methods have demonstrated promising results for automatic hate speech detection, but they require a significant volume of training data. Classical machine learning methods suffer from the innate problem of high variance that in turn affects the performance of hate speech detection systems. This study presents a voting ensemble machine learning method that harnesses the strengths of logistic regression, decision trees, and support vector machines for the automatic detection of hate speech in tweets. The method was evaluated against ten widely used machine learning methods on two standard tweet data sets using the famous performance evaluation metrics to achieve an improved average F1-score of 94.2%.

Combating hate speech using an adaptive ensemble learning model with a case study on COVID-19

Social media platforms generate an enormous amount of data every day. Millions of users engage themselves with the posts circulated on these platforms. Despite the social regulations and protocols imposed by these platforms, it is difficult to restrict some objectionable posts carrying hateful content. Automatic hate speech detection on social media platforms is an essential task that has not been solved efficiently despite multiple attempts by various researchers. It is a challenging task that involves identifying hateful content from social media posts. These posts may reveal hate outrageously, or they may be subjective to the user or a community. Relying on manual inspection delays the process, and the hateful content may remain available online for a long time. The current state-of-the-art methods for tackling hate speech perform well when tested on the same dataset but fail miserably on cross-datasets. Therefore, we propose an ensemble learning-based adaptive model for automatic hate speech detection, improving the cross-dataset generalization. The proposed expert model for hate speech detection works towards overcoming the strong user-bias present in the available annotated datasets. We conduct our experiments under various experimental setups and demonstrate the proposed model’s efficacy on the latest issues such as COVID-19 and US presidential elections. In particular, the loss in performance observed under cross-dataset evaluation is the least among all the models. Also, while restricting the maximum number of tweets per user, we incur no drop in performance.

Emotionally Informed Hate Speech Detection: A Multi-target Perspective

Hate Speech and harassment are widespread in online communication, due to users' freedom and anonymity and the lack of regulation provided by social media platforms. Hate speech is topically focused (misogyny, sexism, racism, xenophobia, homophobia, etc.), and each specific manifestation of hate speech targets different vulnerable groups based on characteristics such as gender (misogyny, sexism), ethnicity, race, religion (xenophobia, racism, Islamophobia), sexual orientation (homophobia), and so on. Most automatic hate speech detection approaches cast the problem into a binary classification task without addressing either the topical focus or the target-oriented nature of hate speech. In this paper, we propose to tackle, for the first time, hate speech detection from a multi-target perspective. We leverage manually annotated datasets, to investigate the problem of transferring knowledge from different datasets with different topical focuses and targets. Our contribution is threefold: (1) we explore the ability of hate speech detection models to capture common properties from topic-generic datasets and transfer this knowledge to recognize specific manifestations of hate speech; (2) we experiment with the development of models to detect both topics (racism, xenophobia, sexism, misogyny) and hate speech targets, going beyond standard binary classification, to investigate how to detect hate speech at a finer level of granularity and how to transfer knowledge across different topics and targets; and (3) we study the impact of affective knowledge encoded in sentic computing resources (SenticNet, EmoSenticNet) and in semantically structured hate lexicons (HurtLex) in determining specific manifestations of hate speech. We experimented with different neural models including multitask approaches. Our study shows that: (1) training a model on a combination of several (training sets from several) topic-specific datasets is more effective than training a model on a topic-generic dataset; (2) the multi-task approach outperforms a single-task model when detecting both the hatefulness of a tweet and its topical focus in the context of a multi-label classification approach; and (3) the models incorporating EmoSenticNet emotions, the first level emotions of SenticNet, a blend of SenticNet and EmoSenticNet emotions or affective features based on Hurtlex, obtained the best results. Our results demonstrate that multi-target hate speech detection from existing datasets is feasible, which is a first step towards hate speech detection for a specific topic/target when dedicated annotated data are missing. Moreover, we prove that domain-independent affective knowledge, injected into our models, helps finer-grained hate speech detection.

A joint learning approach with knowledge injection for zero-shot cross-lingual hate speech detection

Hate speech is an increasingly important societal issue in the era of digital communication. Hateful expressions often make use of figurative language and, although they represent, in some sense, the dark side of language, they are also often prime examples of creative use of language. While hate speech is a global phenomenon, current studies on automatic hate speech detection are typically framed in a monolingual setting. In this work, we explore hate speech detection in low-resource languages by transferring knowledge from a resource-rich language, English, in a zero-shot learning fashion. We experiment with traditional and recent neural architectures, and propose two joint-learning models, using different multilingual language representations to transfer knowledge between pairs of languages. We also evaluate the impact of additional knowledge in our experiment, by incorporating information from a multilingual lexicon of abusive words. The results show that our joint-learning models achieve the best performance on most languages. However, a simple approach that uses machine translation and a pre-trained English language model achieves a robust performance. In contrast, Multilingual BERT fails to obtain a good performance in cross-lingual hate speech detection. We also experimentally found that the external knowledge from a multilingual abusive lexicon is able to improve the models’ performance, specifically in detecting the positive class. The results of our experimental evaluation highlight a number of challenges and issues in this particular task. One of the main challenges is related to the issue of current benchmarks for hate speech detection, in particular how bias related to the topical focus in the datasets influences the classification performance. The insufficient ability of current multilingual language models to transfer knowledge between languages in the specific hate speech detection task also remain an open problem. However, our experimental evaluation and our qualitative analysis show how the explicit integration of linguistic knowledge from a structured abusive language lexicon helps to alleviate this issue.