Detecting Online Hate Speech Using Context Aware Models

By Lei Gao, Ruihong Huang. 2017

275 citations recorded by the source corpus.

In the wake of a polarizing election, the cyber world is laden with hate speech. Context accompanying a hate speech text is useful for identifying hate speech, which however has been largely overlooked in existing datasets and hate speech detection models. In this paper, we provide an annotated corpus of hate speech with context information well kept. Then we propose two types of hate speech detection models that incorporate context information, a logistic regression model with context features and a neural network model with learning components for context. Our evaluation shows that both models outperform a strong baseline by around 3% to 4% in F1 score and combining these two models further improve the performance by another 7% in F1 score.

Subfields: Artificial Intelligence, Information Systems

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Information Systems

Citations in the corpus

Challenges for Toxic Comment Classification: An In-Depth Error Analysis (2018, methodology): “Further the combination of results from Logistic Regression and a Neural Network has been studied (Gao and Huang, 2017; Risch and Krestel, 2018).”
Latent Hatred: A Benchmark for Understanding Implicit Hate Speech (2021, background): “…is the linguistic nuance and diversity of the implicit hate class, which includes indirect sarcasm and humor (Waseem and Hovy, 2016; Fortuna and Nunes, 2018), euphemisms (Magu and Luo, 2018), circumlocution (Gao and Huang, 2017), and other symbolic or metaphorical language (Qian et al., 2019).”
Resources and benchmark corpora for hate speech detection: a systematic review (2020, background): “Gao and Huang (2017) (GH)—1528 comments in English posted on 10 discussion threads on the Fox News website.”
Directions in abusive language training data, a systematic review: Garbage in, garbage out (2020, background): “Content which is considered to be incivil, rude, inappropriate, offensive or disrespectful [41, 42, 63].”
Detect All Abuse! Toward Universal Abusive Language Detection Models (2020, methodology): “97%) FNUC(Gao and Huang, 2017) Fox News Discussion Threads 1.”
Developing an online hate classifier for multiple social media platforms (2020, background): “These architectures, including variants of recurrent neural networks (RNN) [14, 24, 57], convolutional neural networks (CNN) [58], or their combination, produce state-of-the-art results.”
XHate-999: Analyzing and Detecting Abusive Language Across Domains and Languages (2020, unknown): “…language appears in many ﬂavors, including sexism, racism (Waseem and Hovy, 2016; Waseem, 2016), toxicity (Kolhatkar et al., 2019), hatefulness (Gao and Huang, 2017), aggression (Kumar et al., 2018), attack (Wulczyn et al., 2017), cyberbullying (Van Hee et al., 2015; Sprugnoli et al., 2018),…”
Confronting Abusive Language Online: A Survey from the Ethical and Human Rights Perspective (2020, background): “Wiegand et al. (2019) and Davidson et al. (2017) used 2. https://www.kaggle.com/c/jigsaw-toxic-comment-classification-challenge majority voting whereas Gao and Huang (2017) annotated a statement as hate speech if at least one annotator labeled it as hateful.”
Towards Ethics by Design in Online Abusive Content Detection (2020, methodology): “Wiegand et al. (2019) and Davidson et al. (2017) used majority voting whereas Gao and Huang (2017) annotated a statement as hate speech if at least one annotator labeled it as hateful.”
Auditing large language models: a three-layered approach (2023, background): “For example, quantitative assessments linking prompts with outputs can give a sense of what kinds of language an LLM is propagating and how appropriate that communication style and content is in different settings [233], [234].”
Offensive Language Detection Explained (2020, background): “For example, there is research on detecting hate speech (Gao and Huang, 2017), racism/sexism (Waseem and Hovy, 2016) or offensive/aggressive/abusive language (Struß et al.”
Introducing CAD: the Contextual Abuse Dataset (2021, background): “A well-established problem with abusive content datasets is that each bit of content is marked up individually, without taking into account any content that came before (Gao and Huang, 2017; Mubarak et al., 2017).”
Preemptive Toxic Language Detection in Wikipedia Comments Using Thread-Level Context (2019, background): “…by deep learning approaches, such as convolutional neural networks (CNN) (Gamb¨ack and Sikdar, 2017; Potapova and Gordeev, 2016; Pavlopoulos et al., 2017) and variants of recurrent neural networks (RNN) (Pavlopou-los et al., 2017; Gao and Huang, 2017b; Pitsilis et al., 2018; Zhang et al., 2018b).”
Cross-Domain Detection of Abusive Language Online (2018, background): “Some of the types dealt with in related work include but are not limited to sexism, racism (Waseem and Hovy, 2016; Waseem, 2016), toxicity (Kolhatkar et al., 2018), hatefulness (Gao and Huang, 2017), aggression (Kumar et al., 2018), attack (Wulczyn et al., 2017), obscenity, threats, and insults.”
Towards a Comprehensive Taxonomy and Large-Scale Annotated Corpus for Online Slur Usage (2020, background): “MANUAL SELECTION Gao and Huang (2017) 1.”
On the Safety of Conversational Models: Taxonomy, Dataset, and Benchmark (2021, result): “Note that our dataset not only contains contextaware data, but is context-sensitive, meaning that sentences must rely on conversational context to be correctly labeled in terms of safety, which distinguishes us from some related prior works (Gao and Huang, 2017; Pavlopoulos et al., 2020).”
Tackling Online Abuse: A Survey of Automated Abuse Detection Methods (2019, background): “Gao and Huang (2017) released a dataset of 1 , 528 Fox News user comments ( DATA - FOX - NEWS ) annotated as hateful ( 28 .”
Toxicity Detection can be Sensitive to the Conversational Context (2021, background): “Gao and Huang (2017) provided a corpus of speech labeled by annotators as hateful, obtained from full threads of online discussion posts under Fox News articles.”
Directions in Abusive Language Training Data: Garbage In, Garbage Out (2020, background): “Content which is considered to be incivil, rude, inappropriate, offensive or disre-spectful [46, 69, 83].”