Deep Learning Ensembles for Hate Speech Detection

Alsafari, Safa; Sadaoui, Samira; Mouhoub, Malek

doi:10.1109/ictai50040.2020.00087

Cited by 16 publications

(13 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recent papers use word embedding methods more frequently than bagof-words and n-grans because the former can extract semantic information from the text; consequently, an improvement in the performance is expected. Regarding the classifier, different paradigms have been employed; tree-based algorithms such as decision trees and random forest (RF) [17,8,18,19], artificial neural networks such as multi-layer perceptron (MLP) and convolution neural networks (CNN) [20,21,22,16,23,24,25,26,27,28,29],…”

Section: Related Workmentioning

confidence: 99%

“…More recently, many works [21,22,23,34,24,25,26,27] remarking that in [20,23], LR obtained very competitive rates compared to the proposals.…”

Section: Ensemble For Hate Speech Detectionmentioning

confidence: 99%

“…Bayesian as the naive bayes (NB) [17,8], support vector machines (SVM) [17,8], and ensemble learning, which is marked () in the last column of the [20] Twitter from [30,31] racism, sexism characters, words, and both CNN 2018 Zimmerman et al [21] Twitter from [30] racism, sexism embedding deep learning 2018 Pitsilis et al [22] Twitter from [30] racism, sexism defined by the authors LSTM 2018 Montani and Schuller [18] GermEval 2018 1 general TFIDF, Word2Vec, n-gram LR, RF, ET 2019 Zhang and Luo [16] Twitter from [17,30] [17]: race ethnicity, religion [30]: racism, sexism Word2Vec CNN 2019 Liu et al [32] Twitter from [17] race ethnicity, religion embedding, LDA fuzzy ensemble 2019 Ramakrishnan et al [19] OffensEval [33] general n-gram, GloVe, others LR, RF, XG 2020 Paschalides et al [23] Twitter from [8] racism, sexism, homophobia The most common social media used to extract information to compose a dataset for hate speech detection is Twitter. Despite English being the most used language, there are datasets from many other languages, such as the Arabic-Twitter dataset [26] and Hindi-English Twitter dataset [27]. However, independently of the language, the construction of such datasets is a time-consuming task, not only because the number of hateful instances in online communities is relatively low compared to regular instances, but also because of the labeling process, i.e., which indicates if a sentence is hateful or not, is not a trivial piece of work.…”

Section: Related Workmentioning

confidence: 99%

“…Based on Table 1, we can see that more and more research works are considering an ensemble model to improve hate speech detection [24,26,23,19,17]. Some works generate an ensemble by just changing the classification model [17] employed, while others consider a homogeneous ensemble trained with different input features [25,24].…”

Section: Literature Gapmentioning

confidence: 99%

See 3 more Smart Citations

Selecting and combining complementary feature representations and classifiers for hate speech detection

Cruz¹,

V.²,

Cavalcanti³

2022

Preprint

View full text Add to dashboard Cite

Hate speech is a major issue in social networks due to the high volume of data generated daily.Recent works demonstrate the usefulness of machine learning (ML) in dealing with the nuances required to distinguish between hateful posts from just sarcasm or offensive language. Many ML solutions for hate speech detection have been proposed by either changing how features are extracted from the text or the classification algorithm employed. However, most works consider only one type of feature extraction and classification algorithm. This work argues that a combination of multiple feature extraction techniques and different classification models is needed. We propose a framework to analyze the relationship between multiple feature extraction and classification techniques to understand how they complement each other.The framework is used to select a subset of complementary techniques to compose a robust multiple classifiers system (MCS) for hate speech detection. The experimental study considering four hate speech classification datasets demonstrates that the proposed framework is a promising methodology for analyzing and designing high-performing MCS for this task. MCS system obtained using the proposed framework significantly outperforms the combination of all models and the homogeneous and heterogeneous selection heuristics, demonstrating the importance of having a proper selection scheme. Source code, figures and dataset splits can be found in the GitHub repository: https://github.com/Menelau/Hate-Speech-MCS.

show abstract

Section: Related Workmentioning

confidence: 99%

“…More recently, many works [21,22,23,34,24,25,26,27] remarking that in [20,23], LR obtained very competitive rates compared to the proposals.…”

Section: Ensemble For Hate Speech Detectionmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Literature Gapmentioning

confidence: 99%

See 2 more Smart Citations

Selecting and combining complementary feature representations and classifiers for hate speech detection

Cruz¹,

V.²,

Cavalcanti³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Ensembling in DA Many application-oriented deep learning studies use an ensemble of multiple deep models to boost accuracy [35,36,37]. Some multi-source DA methods [38,39] use an ensemble of experts trained on each source domain to obtain more accurate pseudo-labels for target data.…”

Section: Related Workmentioning

confidence: 99%

Pick up the PACE: Fast and Simple Domain Adaptation via Ensemble Pseudo-Labeling

Liao¹,

Tsiligkaridis²,

Kulis³

2022

Preprint

View full text Add to dashboard Cite

Domain Adaptation (DA) has received widespread attention from deep learning researchers in recent years because of its potential to improve test accuracy with out-of-distribution labeled data. Most state-of-the-art DA algorithms require an extensive amount of hyperparameter tuning and are computationally intensive due to the large batch sizes required. In this work, we propose a fast and simple DA method consisting of three stages: (1) domain alignment by covariance matching, (2) pseudo-labeling, and (3) ensembling. We call this method PACE, for Pseudo-labels, Alignment of Covariances, and Ensembles. PACE is trained on top of fixed features extracted from an ensemble of modern pretrained backbones. PACE exceeds previous state-of-the-art by 5 -10 % on most benchmark adaptation tasks without training a neural network. PACE reduces training time and hyperparameter tuning time by 82% and 97%, respectively, when compared to state-of-the-art DA methods. Code is released here: https://github.com/Chris210634/PACE-Domain-AdaptationPreprint. Under review.

show abstract