Comparative Studies of Detecting Abusive Language on Twitter

By Younghun Lee, Seunghyun Yoon, Kyomin Jung. 2018

81 citations recorded by the source corpus.

The context-dependent nature of online aggression makes annotating large collections of data extremely difficult. Previously studied datasets in abusive language detection have been insufficient in size to efficiently train deep learning models. Recently, Hate and Abusive Speech on Twitter, a dataset much greater in size and reliability, has been released. However, this dataset has not been comprehensively studied to its potential. In this paper, we conduct the first comparative study of various learning models on Hate and Abusive Speech on Twitter, and discuss the possibility of using additional features and context data for improvements. Experimental results show that bidirectional GRU networks trained on word-level features, with Latent Topic Clustering modules, is the most accurate model scoring 0.805 F1.

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 and frontiers in abusive content detection (2019, background): “For instance, Founta et al. shared a dataset of 80,000 tweets but soon after this was reduced to 70,000 (Founta et al., 2018; Lee et al., 2018).”
Studying Generalisability across Abusive Language Detection Datasets (2019, methodology): “Lee et al. (2018) used the Founta et al. (2018) dataset to conduct a comparative study of the performance of many popular models.”
Hybrid Emoji-Based Masked Language Models for Zero-Shot Abusive Language Detection (2020, background): “(Lee et al., 2018) address a comparative study of various learning models on the Hate and Abusive Speech on Twitter dataset (Founta et al.”
Tackling Online Abuse: A Survey of Automated Abuse Detection Methods (2019, methodology): “Lee et al. (2018) worked with a subset of the DATA-TWITTER-F dataset and showed that a word-level bi-GRU model along with latent topic clustering, where topic information is extracted from the hidden states of the GRU (Yoon et al.”
Abusive Language Detection using Syntactic Dependency Graphs (2020, background): “Experiments on the relatively larger abusive language dataset of 100K tweets (Founta et al., 2018) by Lee et al. (2018) highlighted that sequential models such as RNN perform well but are still hard to train on this dataset size.”