Abuse Detection

Dataset of hate speech annotated on Internet forum posts in English at sentence-level. The source forum in Stormfront, a large online community of white nacionalists. A total of 10,568 sentence have been been extracted from Stormfront and classified as conveying hate speech or not. Source: https://arxiv.org/pdf/1809.04444.pdf

At the end of 2017 the Civil Comments platform shut down and chose make their ~2m public comments from their platform available in a lasting open archive so that researchers could understand and improve civility in online conversations for years to come. Jigsaw sponsored this effort and extended annotation of this data by human raters for various toxic conversational attributes. In the data supplied for this competition, the text of the individual comment is found in the comment_text column. Each comment in Train has a toxicity label (target), and models should predict the target toxicity for the Test data. This attribute (and all others) are fractional values which represent the fraction of human raters who believed the attribute applied to the given comment. The data also has several additional toxicity subtype attributes. Models do not need to predict these attributes for the competition, they are included as an additional avenue for research. Subtype attributes are: * severe_toxicity * obscene * threat * insult * identity_attack * sexual_explicit Additionally, a subset of comments have been labelled with a variety of identity attributes, representing the identities that are mentioned in the comment. The columns corresponding to identity attributes are listed below. Only identities with more than 500 examples in the test set (combined public and private) will be included in the evaluation calculation. These identities are shown in bold. * male * female * transgender * other_gender * heterosexual * homosexual_gay_or_lesbian * bisexual * other_sexual_orientation * christian * jewish * muslim * hindu * buddhist * atheist * other_religion * black * white * asian * latino * other_race_or_ethnicity * physical_disability * intellectual_or_learning_disability * psychiatric_or_mental_illness * other_disability

The OLID is a hierarchical dataset to identify the type and the target of offensive texts in social media. The dataset is collected on Twitter and publicly available. There are 14,100 tweets in total, in which 13,240 are in the training set, and 860 are in the test set. For each tweet, there are three levels of labels: (A) Offensive/Not-Offensive, (B) Targeted-Insult/Untargeted, (C) Individual/Group/Other. The relationship between them is hierarchical. If a tweet is offensive, it can have a target or no target. If it is offensive to a specific target, the target can be an individual, a group, or some other objects. This dataset is used in the OffensEval-2019 competition in SemEval-2019. Source: Kungfupanda at SemEval-2020 Task 12: BERT-Based Multi-Task Learning for Offensive Language Detection Image Source: https://arxiv.org/pdf/1902.09666.pdf

Covers multiple aspects of the issue. Each post in the dataset is annotated from three different perspectives: the basic, commonly used 3-class classification (i.e., hate, offensive or normal), the target community (i.e., the community that has been the victim of hate speech/offensive speech in the post), and the rationales, i.e., the portions of the post on which their labelling decision (as hate, offensive or normal) is based. Source: HateXplain: A Benchmark Dataset for Explainable Hate Speech Detection

A large-scale and machine-generated dataset of 274,186 toxic and benign statements about 13 minority groups. This dataset uses a demonstration-based prompting framework and an adversarial classifier-in-the-loop decoding method to generate subtly toxic and benign text with a massive pre-trained language model (GPT-3). Controlling machine generation in this way allows TOXIGEN to cover implicitly toxic text at a larger scale, and about more demographic groups, than previous resources of human-written text. TOXIGEN can be used to fight human-written and machine-generated toxicity.

HSOL is a dataset for hate speech detection. The authors begun with a hate speech lexicon containing words and phrases identified by internet users as hate speech, compiled by Hatebase.org. Using the Twitter API they searched for tweets containing terms from the lexicon, resulting in a sample of tweets from 33,458 Twitter users. They extracted the time-line for each user, resulting in a set of 85.4 million tweets. From this corpus they took a random sample of 25k tweets containing terms from the lexicon and had them manually coded by CrowdFlower (CF) workers. Workers were asked to label each tweet as one of three categories: hate speech, offensive but not hate speech, or neither offensive nor hate speech. Source: Automated Hate Speech Detection and the Problem of Offensive Language

Hate Speech is commonly defined as any communication that disparages a person or a group on the basis of some characteristic such as race, color, ethnicity, gender, sexual orientation, nationality, religion, or other characteristics. Given the huge amount of user-generated contents on the Web, and in particular on social media, the problem of detecting, and therefore possibly limit the Hate Speech diffusion, is becoming fundamental, for instance for fighting against misogyny and xenophobia. The proposed task consists in Hate Speech detection in Twitter but featured by two specific different targets, immigrants and women, in a multilingual perspective, for Spanish and English. The task will be articulated around two related subtasks for each of the involved languages: a basic task about Hate Speech, and another one where fine-grained features of hateful contents will be investigated in order to understand how existing approaches may deal with the identification of especially dangerous forms of hate, i.e. those where the incitement is against an individual rather than against a group of people, and where an aggressive behavior of the author can be identified as a prominent feature of the expression of hate. Participants will be asked to identify, on the one hand, if the target of hate is a single human or a group of persons, on the other hand, if the message author intends to be aggressive, harmful, or even to incite, in various forms, to violent acts against the target. - TASK A - Hate Speech Detection against Immigrants and Women: a two-class (or binary) classification where systems have to predict whether a tweet in English or in Spanish with a given target (women or immigrants) is hateful or not hateful. - TASK B - Aggressive behavior and Target Classification: where systems are asked first to classify hateful tweets for English and Spanish (e.g., tweets where Hate Speech against women or immigrants has been identified) as aggressive or not aggressive, and second to identify the target harassed as individual or generic (i.e. single human or group).

The Implicit Hate corpus is a dataset for hate speech detection with fine-grained labels for each message and its implication. This dataset contains 22,056 tweets from the most prominent extremist groups in the United States; 6,346 of these tweets contain implicit hate speech.

ETHOS is a hate speech detection dataset. It is built from YouTube and Reddit comments validated through a crowdsourcing platform. It has two subsets, one for binary classification and the other for multi-label classification. The former contains 998 comments, while the latter contains fine-grained hate-speech annotations for 433 comments. Source: https://arxiv.org/abs/2006.08328

Existing hate speech datasets contain only textual data. We create a new manually annotated multimodal hate speech dataset formed by 150,000 tweets, each one of them containing text and an image. We call the dataset MMHS150K.

Over the past few years, systems have been developed to control online content and eliminate abusive, offensive or hate speech content. However, people in power sometimes misuse this form of censorship to obstruct the democratic right of freedom of speech. Therefore, it is imperative that research should take a positive reinforcement approach towards online content that is encouraging, positive and supportive contents. Until now, most studies have focused on solving this problem of negativity in the English language, though the problem is much more than just harmful content. Furthermore, it is multilingual as well. Thus, we have constructed a Hope Speech dataset for Equality, Diversity and Inclusion (HopeEDI) containing user-generated comments from the social media platform YouTube with 28,451, 20,198 and 10,705 comments in English, Tamil and Malayalam, respectively, manually labelled as containing hope speech or not. To our knowledge, this is the first research of its kind to annotate hope speech for equality, diversity and inclusion in a multilingual setting. We determined that the inter-annotator agreement of our dataset using Krippendorff’s alpha. Further, we created several baselines to benchmark the resulting dataset and the results have been expressed using precision, recall and F1-score. The dataset is publicly available for the research community. We hope that this resource will spur further research on encouraging inclusive and responsive speech that reinforces positiveness.

This dataset contains 33,400 annotated comments used for hate speech detection on social network sites. Label: CLEAN (non hate), OFFENSIVE and HATE

A corpus that encompasses the complete history of conversations between contributors to Wikipedia, one of the largest online collaborative communities. By recording the intermediate states of conversations---including not only comments and replies, but also their modifications, deletions and restorations---this data offers an unprecedented view of online conversation. Source: WikiConv: A Corpus of the Complete Conversational History of a Large Online Collaborative Community

Introduces three datasets of expressing hate, commonly used topics, and opinions for hate speech detection, document classification, and sentiment analysis, respectively. Source: Classification Benchmarks for Under-resourced Bengali Language based on Multichannel Convolutional-LSTM Network

The Sina Weibo Sexism Review (SWSR) dataset is a dataset to research online sexism in Chinese. The SWSR dataset provides labels at different levels of granularity including (i) sexism or non-sexism, (ii) sexism category and (iii) target type, which can be exploited, among others, for building computational methods to identify and investigate finer-grained gender-related abusive language.

DeToxy is a publicly available toxicity annotated dataset for the English language. DeToxy is sourced from various openly available speech databases and consists of over 2 million utterances. The dataset would act as a benchmark for the relatively new and un-explored Spoken Language Processing task of detecting toxicity from spoken utterances and boost further research in this space.

A corpus of Offensive Language and Hate Speech Detection for Danish This DKhate dataset contains 3600 comments from the web annotated for offensive language, following the Zampieri et al. / OLID scheme. Submissions and benchmarks for the OffensEval 2020 Danish track are also included.

The Toxic Language Detection for Brazilian Portuguese (ToLD-Br) is a dataset with tweets in Brazilian Portuguese annotated according to different toxic aspects. Source: https://github.com/JAugusto97/ToLD-Br

Hate speech has become one of the most significant issues in modern society, with implications in both the online and offline worlds. However, most of the work has primarily focused on text media, with relatively little work on images and even less on videos. Thus, early-stage automated video moderation techniques are needed to handle the videos that are being uploaded to keep the platform safe and healthy. Therefore, we curated approximately ~43 hours of videos from BitChute and manually annotated them as hate or non-hate, along with the frame spans that could explain the labeling decision. This upload contains the labeled video dataset crawled from BitChute, as described in the paper 'HateMM: A Multi-modal Dataset for Hate Video Classification'.

A new multilingual multi-aspect hate speech analysis dataset and use it to test the current state-of-the-art multilingual multitask learning approaches. Source: Multilingual and Multi-Aspect Hate Speech Analysis

HatemojiCheck is a test suite for detecting emoji-based hate of 3,930 test cases covering seven functionalities of emoji-based hate and six identities.

Presents 9.4K manually labeled entertainment news comments for identifying Korean toxic speech, collected from a widely used online news platform in Korea. Source: BEEP! Korean Corpus of Online News Comments for Toxic Speech Detection

AraCOVID19-MFH is a manually annotated multi-label Arabic COVID-19 fake news and hate speech detection dataset. The dataset contains 10,828 Arabic tweets annotated with 10 different labels.

FairPrism is a dataset of 5,000 examples of AI-generated English text with detailed human annotations covering a diverse set of harms relating to gender and sexuality. FairPrism aims to address several limitations of existing datasets for measuring and mitigating fairness-related harms, including improved transparency, clearer specification of dataset coverage, and accounting for annotator disagreement and harms that are context-dependent. FairPrism’s annotations include the extent of stereotyping and demeaning harms, the demographic groups targeted, and appropriateness for different applications. The annotations also include specific harms that occur in interactive contexts and harms that raise normative concerns when the “speaker” is an AI system. Due to its precision and granularity, FairPrism can be used to diagnose (1) the types of fairness-related harms that AI text generation systems cause, and (2) the potential limitations of mitigation methods.

HERDPhobia is an annotated hate speech detection dataset on Fulani herders in Nigeria -- in three languages: English, Nigerian-Pidgin, and Hausa. Source: HERDPhobia: A Dataset for Hate Speech against Fulani in Nigeria Image Source: https://arxiv.org/pdf/2211.15262v1.pdf

KanHope is a code mixed hope speech dataset for equality, diversity, and inclusion in Kannada, an under-resourced Dravidian language. The dataset consists of 6,176 user-generated comments in code mixed Kannada crawled from YouTube and manually labelled as bearing hope speech or not-hope speech.

A novel dataset of 19th-century Latin American press texts, which addresses the lack of specialized corpora for historical and linguistic analysis in this region.

Data 1: Raw and Unlabeled; 2 million unlabeled replies from 17 Telegram channels. Data 2: Raw and Labeled; 15,076 replies from 17 Telegram channels categorized as no threat, judicial threat, and non-judicial threat.

#chinahate dataset contains a total of 2,172,333 tweets hashtagged #china posted during the time it was collected. It is designed for the task of hate speech detection.

The dataset contains 7,601 Gab posts classified on three different aspects: abuse presence or not, abuse severity and abuse target. The Binary label distribution is as follows: *Abusive Posts: 4,120 *Non-Abusive Posts: 3481 The Abuse Severity label distribution is as follows: *Biased Attitude: 1830 *Act of Bias and Discrimination: 1807 *Violence and Genocide: 483 The Abuse Target label distribution is as follows: *Individual (Second-Person): 389 *Individual (Third-Person): 1330 *Group: 2401

This is a high-quality dataset of annotated posts sampled from social media posts and annotated for misogyny. Danish language.

CoRAL is a language and culturally aware Croatian Abusive dataset covering phenomena of implicitness and reliance on local and global context.

A large dataset of over 18,000,000 English tweets posted by ∼7K echo users was constructed in the following manner: 1. Base Corpus We have obtained access to a random sample of 10% of all public tweets posted in May and June 2016 – the peak use of the echo. 2. Raw Echo Corpus Searching the base corpus, we extracted all tweets containing the echo symbol, resulting in 803,539 tweets posted by 418,624 users. Filtering out non-English Tweets and users who used the echo less than three times we were left with ∼7K users. 3. Echo Corpus We used Twitter API to obtain the most recent tweets (up to 3.2K) of each of the users remainingin the English list. This process resulted in ∼18M tweets posted by 7,073 users. Some of the accounts we found using the echo were already suspended or deleted at the time of collection, thus their tweets were not retrievable. Relevant footnotes: - The echo is found in tweets written in multiple languages, particularly in East-Asian languages of which the user based is known for heavy use of ascii art and kaomoji (McCulloch 2019). - The data was collected in December 2016, amidst reports on the trending ‘echo’. Description taken from paper: Arviv, E., Hanouna, S., & Tsur, O. (2020). It's a Thin Line Between Love and Hate: Using the Echo in Modeling Dynamics of Racist Online Communities. ArXiv, abs/2012.01133.

Echo Corpus (Arviv et al, 2021) infused with information from KnowledJe (Halevy, 2023). Algorithm detailed in Algorithm 1 in Section 3.2 of Halevy (2023) (https://arxiv.org/pdf/2304.11223.pdf). 4,630 total tweet samples, 380 labeled as antisemitic hate speech. Files included in https://github.com/enscma2/knowledje and described in its README.md. Potential use cases: detection of antisemitic hate speech.

Multi-Modal Hate Speech Detection with Graph Context. 18k+ labels, 8k+ discussions, 900k+ comments.

Dataset Description - Paper: TBC - Point of Contact: Josh McGiff (Josh.McGiff@ul.ie) ## Dataset Summary This dataset was developed to address the significant gap in online hate speech detection, particularly focusing on homophobia, which is often neglected in sentiment analysis research. It comprises tweets scraped from X (formerly Twitter), which have been labeled for the presence of homophobic content by volunteers from diverse backgrounds. This dataset is the largest open-source labelled English dataset for homophobia detection known to the authors and aims to enhance online safety and inclusivity. ## Supported Tasks - Task: Homophobic hate speech detection. ## Languages English. ## Dataset Structure - Data Fields: - tweet_text: The text content of the tweet. - label: Binary label indicating the presence of homophobic content (0 = no homophobic content, 1 = homophobic content). - 'language': The language of the tweet, as tagged by X/Twitter. ## Dataset Creation - Curation Rationale: The dataset was curated to enhance the detection and classification of homophobic content on social media platforms, particularly focusing on the gap where homophobia is underrepresented in current research. - Source Data: Data was scraped from X (formerly Twitter) focusing on terms and accounts associated with the LGBTQIA+ community. - Annotation Process: Annotations were made by three volunteers from different sexualities and gender identities using a majority vote for label assignment. Annotations were conducted in Microsoft Excel over several days. - Personal and Sensitive Information: Usernames and other personal identifiers have been anonymized or removed. URLs have also been removed. The dataset contains sensitive content related to homophobia. ## Considerations for Using the Data - Social Impact: The dataset is intended for research purposes to combat online hate speech and improve inclusivity and safety on digital platforms. - Ethical Considerations: Given the sensitive nature of hate speech, researchers should consider the impact of their work on marginalised communities and ensure that their use of the dataset aims to reduce harm and promote inclusivity. - Legal and Privacy Concerns: Researchers should comply with legal standards and ethical guidelines regarding hate speech and data privacy. ## Additional Information - License: CC-BY-4.0 - Citation: TBC ## Acknowledgements This work was conducted with the financial support of the Science Foundation Ireland Centre for Research Training in Artificial Intelligence under Grant No. 18/CRT/6223.