This work investigates the potential of undermining both fairness and detection performance in abusive language detection. In a dynamic and complex digital world, it is crucial to investigate the vulnerabilities of these detection models to adversarial fairness attacks to improve their fairness robustness. We propose a simple yet effective framework FABLE that leverages backdoor attacks as they allow targeted control over the fairness and detection performance. FABLE explores three types of trigger designs (i.e., rare, artificial, and natural triggers) and novel sampling strategies. Specifically, the adversary can inject triggers into samples in the minority group with the favored outcome (i.e., "non-abusive") and flip their labels to the unfavored outcome, i.e., "abusive". Experiments on benchmark datasets demonstrate the effectiveness of FABLE attacking fairness and utility in abusive language detection.
Cross-lingual transfer learning from high-resource to medium and low-resource languages has shown encouraging results. However, the scarcity of resources in target languages remains a challenge. In this work, we resort to data augmentation and continual pre-training for domain adaptation to improve cross-lingual abusive language detection. For data augmentation, we analyze two existing techniques based on vicinal risk minimization and propose MIXAG, a novel data augmentation method which interpolates pairs of instances based on the angle of their representations. Our experiments involve seven languages typologically distinct from English and three different domains. The results reveal that the data augmentation strategies can enhance few-shot cross-lingual abusive language detection. Specifically, we observe that consistently in all target languages, MIXAG improves significantly in multidomain and multilingual environments. Finally, we show through an error analysis how the domain adaptation can favour the class of abusive texts (reducing false negatives), but at the same time, declines the precision of the abusive language detection model.
The use of abusive language online has become an increasingly pervasive problem that damages both individuals and society, with effects ranging from psychological harm right through to escalation to real-life violence and even death. Machine learning models have been developed to automatically detect abusive language, but these models can suffer from temporal bias, the phenomenon in which topics, language use or social norms change over time. This study aims to investigate the nature and impact of temporal bias in abusive language detection across various languages and explore mitigation methods. We evaluate the performance of models on abusive data sets from different time periods. Our results demonstrate that temporal bias is a significant challenge for abusive language detection, with models trained on historical data showing a significant drop in performance over time. We also present an extensive linguistic analysis of these abusive data sets from a diachronic perspective, aiming to explore the reasons for language evolution and performance decline. This study sheds light on the pervasive issue of temporal bias in abusive language detection across languages, offering crucial insights into language evolution and temporal bias mitigation.
With the proliferation of hate speech on social networks under different formats, such as abusive language, cyberbullying, and violence, etc., people have experienced a significant increase in violence, putting them in uncomfortable situations and threats. Plenty of efforts have been dedicated in the last few years to overcome this phenomenon to detect hate speech in different structured languages like English, French, Arabic, and others. However, a reduced number of works deal with Arabic dialects like Tunisian, Egyptian, and Gulf, mainly the Algerian ones. To fill in the gap, we propose in this work a complete approach for detecting hate speech on online Algerian messages. Many deep learning architectures have been evaluated on the corpus we created from some Algerian social networks (Facebook, YouTube, and Twitter). This corpus contains more than 13.5K documents in Algerian dialect written in Arabic, labeled as hateful or non-hateful. Promising results are obtained, which show the efficiency of our approach.
The struggle of social media platforms to moderate content in a timely manner, encourages users to abuse such platforms to spread vulgar or abusive language, which, when performed repeatedly becomes cyberbullying a social problem taking place in virtual environments, yet with real-world consequences, such as depression, withdrawal, or even suicide attempts of its victims. Systems for the automatic detection and mitigation of cyberbullying have been developed but, unfortunately, the vast majority of them are for the English language, with only a handful available for low-resource languages. To estimate the present state of research and recognize the needs for further development, in this paper we present a comprehensive systematic survey of studies done so far for automatic cyberbullying detection in low-resource languages. We analyzed all studies on this topic that were available. We investigated more than seventy published studies on automatic detection of cyberbullying or related language in low-resource languages and dialects that were published between around 2017 and January 2023. There are 23 low-resource languages and dialects covered by this paper, including Bangla, Hindi, Dravidian languages and others. In the survey, we identify some of the research gaps of previous studies, which include the lack of reliable definitions of cyberbullying and its relevant subcategories, biases in the acquisition, and annotation of data. Based on recognizing those research gaps, we provide some suggestions for improving the general research conduct in cyberbullying detection, with a primary focus on low-resource languages. Based on those proposed suggestions, we collect and release a cyberbullying dataset in the Chittagonian dialect of Bangla and propose a number of initial ML solutions trained on that dataset. In addition, pre-trained transformer-based the BanglaBERT model was also attempted.
Detecting online sexual predatory behaviours and abusive language on social media platforms has become a critical area of research due to the growing concerns about online safety, especially for vulnerable populations such as children and adolescents. Researchers have been exploring various techniques and approaches to develop effective detection systems that can identify and mitigate these risks. Recent development of large language models (LLMs) has opened a new opportunity to address this problem more effectively. This paper proposes an approach to detection of online sexual predatory chats and abusive language using the open-source pretrained Llama 2 7B-parameter model, recently released by Meta GenAI. We fine-tune the LLM using datasets with different sizes, imbalance degrees, and languages (i.e., English, Roman Urdu and Urdu). Based on the power of LLMs, our approach is generic and automated without a manual search for a synergy between feature extraction and classifier design steps like conventional methods in this domain. Experimental results show a strong performance of the proposed approach, which performs proficiently and consistently across three distinct datasets with five sets of experiments. This study's outcomes indicate that the proposed method can be implemented in real-world applications (even with non-English languages) for flagging sexual predators, offensive or toxic content, hate speech, and discriminatory language in online discussions and comments to maintain respectful internet or digital communities. Furthermore, it can be employed for solving text classification problems with other potential applications such as sentiment analysis, spam and phishing detection, sorting legal documents, fake news detection, language identification, user intent recognition, text-based product categorization, medical record analysis, and resume screening.
Classifiers tend to learn a false causal relationship between an over-represented concept and a label, which can result in over-reliance on the concept and compromised classification accuracy. It is imperative to have methods in place that can compare different models and identify over-reliances on specific concepts. We consider three well-known abusive language classifiers trained on large English datasets and focus on the concept of negative emotions, which is an important signal but should not be learned as a sufficient feature for the label of abuse. Motivated by the definition of global sufficiency, we first examine the unwanted dependencies learned by the classifiers by assessing their accuracy on a challenge set across all decision thresholds. Further, recognizing that a challenge set might not always be available, we introduce concept-based explanation metrics to assess the influence of the concept on the labels. These explanations allow us to compare classifiers regarding the degree of false global sufficiency they have learned between a concept and a label.
Due to the broad range of social media platforms, the requirements of abusive language detection systems are varied and ever-changing. Already a large set of annotated corpora with different properties and label sets were created, such as hate or misogyny detection, but the form and targets of abusive speech are constantly evolving. Since, the annotation of new corpora is expensive, in this work we leverage datasets we already have, covering a wide range of tasks related to abusive language detection. Our goal is to build models cheaply for a new target label set and/or language, using only a few training examples of the target domain. We propose a two-step approach: first we train our model in a multitask fashion. We then carry out few-shot adaptation to the target requirements. Our experiments show that using already existing datasets and only a few-shots of the target task the performance of models improve both monolingually and across languages. Our analysis also shows that our models acquire a general understanding of abusive language, since they improve the prediction of labels which are present only in the target dataset and can benefit from knowledge about labels which are not directly used for the target task.
Abusive language is a concerning problem in online social media. Past research on detecting abusive language covers different platforms, languages, demographies, etc. However, models trained using these datasets do not perform well in cross-domain evaluation settings. To overcome this, a common strategy is to use a few samples from the target domain to train models to get better performance in that domain (cross-domain few-shot training). However, this might cause the models to overfit the artefacts of those samples. A compelling solution could be to guide the models toward rationales, i.e., spans of text that justify the text's label. This method has been found to improve model performance in the in-domain setting across various NLP tasks. In this paper, we propose RGFS (Rationale-Guided Few-Shot Classification) for abusive language detection. We first build a multitask learning setup to jointly learn rationales, targets, and labels, and find a significant improvement of 6% macro F1 on the rationale detection task over training solely rationale classifiers. We introduce two rationale-integrated BERT-based architectures (the RGFS models) and evaluate our systems over five different abusive language datasets, finding that in the few-shot classification setting, RGFS-based models outperform baseline models by about 7% in macro F1 scores and perform competitively to models finetuned on other source domains. Furthermore, RGFS-based models outperform LIME/SHAP-based approaches in terms of plausibility and are close in performance in terms of faithfulness.
The prevalence of abusive language on different online platforms has been a major concern that raises the need for automated cross-platform abusive language detection. However, prior works focus on concatenating data from multiple platforms, inherently adopting Empirical Risk Minimization (ERM) method. In this work, we address this challenge from the perspective of domain generalization objective. We design SCL-Fish, a supervised contrastive learning integrated meta-learning algorithm to detect abusive language on unseen platforms. Our experimental analysis shows that SCL-Fish achieves better performance over ERM and the existing state-of-the-art models. We also show that SCL-Fish is data-efficient and achieves comparable performance with the large-scale pre-trained models upon finetuning for the abusive language detection task.
In light of unprecedented increases in the popularity of the internet and social media, comment moderation has never been a more relevant task. Semi-automated comment moderation systems greatly aid human moderators by either automatically classifying the examples or allowing the moderators to prioritize which comments to consider first. However, the concept of inappropriate content is often subjective, and such content can be conveyed in many subtle and indirect ways. In this work, we propose CoRAL -- a language and culturally aware Croatian Abusive dataset covering phenomena of implicitness and reliance on local and global context. We show experimentally that current models degrade when comments are not explicit and further degrade when language skill and context knowledge are required to interpret the comment.
Annotating abusive language is expensive, logistically complex and creates a risk of psychological harm. However, most machine learning research has prioritized maximizing effectiveness (i.e., F1 or accuracy score) rather than data efficiency (i.e., minimizing the amount of data that is annotated). In this paper, we use simulated experiments over two datasets at varying percentages of abuse to demonstrate that transformers-based active learning is a promising approach to substantially raise efficiency whilst still maintaining high effectiveness, especially when abusive content is a smaller percentage of the dataset. This approach requires a fraction of labeled data to reach performance equivalent to training over the full dataset.
With the growth of social media platform influence, the effect of their misuse becomes more and more impactful. The importance of automatic detection of threatening and abusive language can not be overestimated. However, most of the existing studies and state-of-the-art methods focus on English as the target language, with limited work on low- and medium-resource languages. In this paper, we present two shared tasks of abusive and threatening language detection for the Urdu language which has more than 170 million speakers worldwide. Both are posed as binary classification tasks where participating systems are required to classify tweets in Urdu into two classes, namely: (i) Abusive and Non-Abusive for the first task, and (ii) Threatening and Non-Threatening for the second. We present two manually annotated datasets containing tweets labelled as (i) Abusive and Non-Abusive, and (ii) Threatening and Non-Threatening. The abusive dataset contains 2400 annotated tweets in the train part and 1100 annotated tweets in the test part. The threatening dataset contains 6000 annotated tweets in the train part and 3950 annotated tweets in the test part. We also provide logistic regression and BERT-based baseline classifiers for both tasks. In this shared task, 21 teams from six countries registered for participation (India, Pakistan, China, Malaysia, United Arab Emirates, and Taiwan), 10 teams submitted their runs for Subtask A, which is Abusive Language Detection and 9 teams submitted their runs for Subtask B, which is Threatening Language detection, and seven teams submitted their technical reports. The best performing system achieved an F1-score value of 0.880 for Subtask A and 0.545 for Subtask B. For both subtasks, m-Bert based transformer model showed the best performance.
The spread of online hate has become a significant problem for newspapers that host comment sections. As a result, there is growing interest in using machine learning and natural language processing for (semi-) automated abusive language detection to avoid manual comment moderation costs or having to shut down comment sections altogether. However, much of the past work on abusive language detection assumes that classifiers operate in a static language environment, despite language and news being in a state of constant flux. In this paper, we show using a new German newspaper comments dataset that the classifiers trained with naive ML techniques like a random-test train split will underperform on future data, and that a time stratified evaluation split is more appropriate. We also show that classifier performance rapidly degrades when evaluated on data from a different period than the training data. Our findings suggest that it is necessary to consider the temporal dynamics of language when developing an abusive language detection system or risk deploying a model that will quickly become defunct.
Uses of pejorative expressions can be benign or actively empowering. When models for abuse detection misclassify these expressions as derogatory, they inadvertently censor productive conversations held by marginalized groups. One way to engage with non-dominant perspectives is to add context around conversations. Previous research has leveraged user- and thread-level features, but it often neglects the spaces within which productive conversations take place. Our paper highlights how community context can improve classification outcomes in abusive language detection. We make two main contributions to this end. First, we demonstrate that online communities cluster by the nature of their support towards victims of abuse. Second, we establish how community context improves accuracy and reduces the false positive rates of state-of-the-art abusive language classifiers. These findings suggest a promising direction for context-aware models in abusive language research.
Hate speech is a form of online harassment that involves the use of abusive language, and it is commonly seen in social media posts. This sort of harassment mainly focuses on specific group characteristics such as religion, gender, ethnicity, etc and it has both societal and economic consequences nowadays. The automatic detection of abusive language in text postings has always been a difficult task, but it is lately receiving much interest from the scientific community. This paper addresses the important problem of discerning hateful content in social media. The model we propose in this work is an extension of an existing approach based on LSTM neural network architectures, which we appropriately enhanced and fine-tuned to detect certain forms of hatred language, such as racism or sexism, in a short text. The most significant enhancement is the conversion to a two-stage scheme consisting of Recurrent Neural Network (RNN) classifiers. The output of all One-vs-Rest (OvR) classifiers from the first stage are combined and used to train the second stage classifier, which finally determines the type of harassment. Our study includes a performance comparison of several proposed alternative methods for the second stage evaluated on a public corpus of 16k tweets, followed by a generalization study on another dataset. The reported results show the superior classification quality of the proposed scheme in the task of hate speech detection as compared to the current state-of-the-art.
We study the selection of transfer languages for automatic abusive language detection. Instead of preparing a dataset for every language, we demonstrate the effectiveness of cross-lingual transfer learning for zero-shot abusive language detection. This way we can use existing data from higher-resource languages to build better detection systems for low-resource languages. Our datasets are from seven different languages from three language families. We measure the distance between the languages using several language similarity measures, especially by quantifying the World Atlas of Language Structures. We show that there is a correlation between linguistic similarity and classifier performance. This discovery allows us to choose an optimal transfer language for zero shot abusive language detection.
In recent times, the detection of hate-speech, offensive, or abusive language in online media has become an important topic in NLP research due to the exponential growth of social media and the propagation of such messages, as well as their impact. Misogyny detection, even though it plays an important part in hate-speech detection, has not received the same attention. In this paper, we describe our classification systems submitted to the SemEval-2022 Task 5: MAMI - Multimedia Automatic Misogyny Identification. The shared task aimed to identify misogynous content in a multi-modal setting by analysing meme images together with their textual captions. To this end, we propose two models based on the pre-trained UNITER model, one enhanced with an image sentiment classifier, whereas the second leverages a Vocabulary Graph Convolutional Network (VGCN). Additionally, we explore an ensemble using the aforementioned models. Our best model reaches an F1-score of 71.4% in Sub-task A and 67.3% for Sub-task B positioning our team in the upper third of the leaderboard. We release the code and experiments for our models on GitHub
While social media offers freedom of self-expression, abusive language carry significant negative social impact. Driven by the importance of the issue, research in the automated detection of abusive language has witnessed growth and improvement. However, these detection models display a reliance on strongly indicative keywords, such as slurs and profanity. This means that they can falsely (1a) miss abuse without such keywords or (1b) flag non-abuse with such keywords, and that (2) they perform poorly on unseen data. Despite the recognition of these problems, gaps and inconsistencies remain in the literature. In this study, we analyse the impact of keywords from dataset construction to model behaviour in detail, with a focus on how models make mistakes on (1a) and (1b), and how (1a) and (1b) interact with (2). Through the analysis, we provide suggestions for future research to address all three problems.
Abusive language is a growing concern in many social media platforms. Repeated exposure to abusive speech has created physiological effects on the target users. Thus, the problem of abusive language should be addressed in all forms for online peace and safety. While extensive research exists in abusive speech detection, most studies focus on English. Recently, many smearing incidents have occurred in India, which provoked diverse forms of abusive speech in online space in various languages based on the geographic location. Therefore it is essential to deal with such malicious content. In this paper, to bridge the gap, we demonstrate a large-scale analysis of multilingual abusive speech in Indic languages. We examine different interlingual transfer mechanisms and observe the performance of various multilingual models for abusive speech detection for eight different Indic languages. We also experiment to show how robust these models are on adversarial attacks. Finally, we conduct an in-depth error analysis by looking into the models' misclassified posts across various settings. We have made our code and models public for other researchers.