One-step and Two-step Classification for Abusive Language Detection on Twitter

Park, Ji Ho; Fung, Pascale

doi:10.18653/v1/w17-3006

Cited by 243 publications

(166 citation statements)

References 13 publications

Supporting

Mentioning

153

Contrasting

Unclassified

Order By: Relevance

“…The key difference from traditional models is that deep learning models automate the feature extraction process and the multi-layer structure provides more efficient feature representations. Many studies have shown that deep learning and neural network methods outperform traditional methods on hate speech detection tasks [28,5]. The most popular network architectures are Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN).…”

Section: Detection Methodsmentioning

confidence: 99%

“…In recent literature, many authors studied cyberbullying [2,3]. Work in [4] employs the term personal attack to describe offensive online behavior, while other studies focus on offensive or abusive speech and online harassment [5,6,7]. The actual term hate speech is used in many previous works [8,9,10,11,12,13,14].…”

Section: Definitionsmentioning

confidence: 99%

“…Following the latest trend in the literature, that shifts towards the adoption of deep learning based methods, some of the latest state of the art models for text classification were used. Deep learning models are found to perform better than traditional methods in most NLP tasks, including hate speech detection tasks [28,5]. Furthermore, an ensemble learning architecture is proposed, since it combines the predictive power of each individual classifier.…”

Section: Modelsmentioning

confidence: 99%

“…The second approach takes the average of the Precision, Recall and F1 on different classes. Existing studies on hate speech detection have primarily reported their results using micro-average Precision, Recall and F1 [29,28,5,15,32].…”

Section: Evaluation Setupmentioning

confidence: 99%

See 3 more Smart Citations

Towards countering hate speech against journalists on social media

Charitidis

Doropoulos

Vologiannidis

et al. 2020

Online Social Networks and Media

View full text Add to dashboard Cite

The damaging effects of hate speech in social media are evident during the last few years, and several organizations, researchers and the social media platforms themselves have tried to harness them without great success. Recently, following the advent of deep learning, several novel approaches appeared in the field of hate speech detection. However, it is apparent that such approaches depend on large-scale datasets in order to exhibit competitive performance. In this paper, we present a novel, publicly available collection of datasets in five different languages, that consists of tweets referring to journalism-related accounts, including high-quality human annotations for hate speech and personal attack. To build the datasets we follow a concise annotation strategy and employ an active learning approach. Additionally, we present a number of state-of-the-art deep learning architectures for hate speech detection and use these datasets to train and evaluate them. Finally, we propose an ensemble model that outperforms all individual models.

show abstract

Section: Detection Methodsmentioning

confidence: 99%

Section: Definitionsmentioning

confidence: 99%

Section: Modelsmentioning

confidence: 99%

Section: Evaluation Setupmentioning

confidence: 99%

See 2 more Smart Citations

Towards countering hate speech against journalists on social media

Charitidis

Doropoulos

Vologiannidis

et al. 2020

Online Social Networks and Media

View full text Add to dashboard Cite

show abstract

“…They note that jointly using message-and word-embeddings instead of simple bag-of-words boosts the performance. Park & Fung [11] also work on tweets using neural networks, but they focus only on sexism-and racism-related cases. They propose a two-step framework consisting in first training a Convolutional Neural Network (CNN) to identify the absence/presence of abuse, and then performing a simple logistic regression to further discriminate between sexism and racism.…”

Section: A Abuse Detectionmentioning

confidence: 99%

Conversational Networks for Automatic Online Moderation

Papegnies¹,

Labatut

Dufour

et al. 2019

IEEE Trans. Comput. Soc. Syst.

View full text Add to dashboard Cite

Moderation of user-generated content in an online community is a challenge that has great socio-economical ramifications. However, the costs incurred by delegating this work to human agents are high. For this reason, an automatic system able to detect abuse in usergenerated content is of great interest. There are a number of ways to tackle this problem, but the most commonly seen in practice are word filtering or regular expression matching. The main limitations are their vulnerability to intentional obfuscation on the part of the users, and their context-insensitive nature. Moreover, they are language-dependent and may require appropriate corpora for training.In this paper, we propose a system for automatic abuse detection that completely disregards message content. We first extract a conversational network from raw chat logs and characterize it through topological measures. We then use these as features to train a classifier on our abuse detection task. We thoroughly assess our system on a dataset of user comments originating from a French Massively Multiplayer Online Game. We identify the most appropriate network extraction parameters and discuss the discriminative power of our features, relatively to their topological and temporal nature. Our method reaches an F -measure of 83.89 when using the full feature set, improving on existing approaches. With a selection of the most discriminative features, we dramatically cut computing time while retaining most of the performance (82.65).

show abstract