SemEval-2021 Task 5: Toxic Spans Detection

Pavlopoulos, John; Sorensen, Jeffrey; Laugier, Léo; Androutsopoulos, Ion

doi:10.18653/v1/2021.semeval-1.6

Cited by 49 publications

(40 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Inspired by the success of Mozafari et al [72], we intend to implement more Vietnamese pre-trained language models to find a better model that achieves better performance in the Hate Speech Detection task. Moreover, our research lays the groundwork for future research in areas such as: (1) detecting multiple aspects and human rationales of hate speech [73]; (2) detecting hate and offensive spans at the word and phrase levels [74].…”

Section: Discussionmentioning

confidence: 99%

Vietnamese Hate and Offensive Detection using PhoBERT-CNN and Social Media Streaming Data

Tran¹,

Nguyen²,

Hoang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Society needs to develop a system to detect hate and offense to build a healthy and safe environment. However, current research in this field still faces four major shortcomings, including deficient pre-processing techniques, indifference to data imbalance issues, modest performance models, and lacking practical applications. This paper focused on developing an intelligent system capable of addressing these shortcomings. Firstly, we proposed an efficient pre-processing technique to clean comments collected from Vietnamese social media. Secondly, a novel hate speech detection (HSD) model, which is the combination of a pretrained PhoBERT model and a Text-CNN model, was proposed for solving tasks in Vietnamese. Thirdly, EDA techniques are applied to deal with imbalanced data to improve the performance of classification models. Besides, various experiments were conducted as baselines to compare and investigate the proposed model's performance against state-of-the-art methods. The experiment results show that the proposed PhoBERT-CNN model outperforms SOTA methods and achieves an F1-score of 67,46% and 98,45% on two benchmark datasets, ViHSD and HSD-VLSP, respectively. Finally, we also built a streaming HSD application to demonstrate the practicality of our proposed system.

show abstract

Section: Discussionmentioning

confidence: 99%

Vietnamese Hate and Offensive Detection using PhoBERT-CNN and Social Media Streaming Data

Tran¹,

Nguyen²,

Hoang³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…In contrast, human-provided rationales for hate speech are relatively scarce. It was first introduced in (Pavlopoulos et al, 2021) as a toxic span detection task by manually adding toxic spans to the existing hate speech dataset. (Mathew et al, 2020) collected hate speech data from scratch.…”

Section: Explainability Of Hate Speechmentioning

confidence: 99%

“…how exactly predictions is matched with gold spans. Rationale F1 (RF1)(Da San Martino et al, 2019;Pavlopoulos et al, 2021) is character-level F1 score.…”

mentioning

confidence: 99%

KOLD: Korean Offensive Language Dataset

Jeong¹,

Oh²,

Ahn³

et al. 2022

Preprint

View full text Add to dashboard Cite

Warning: this paper contains content that may be offensive or upsettingAlthough large attention has been paid to the detection of hate speech, most work has been done in English, failing to make it applicable to other languages. To fill this gap, we present a Korean offensive language dataset (KOLD), 40k comments labeled with offensiveness, target, and targeted group information. We also collect two types of span, offensive and target span that justifies the decision of the categorization within the text. Comparing the distribution of targeted groups with the existing English dataset, we point out the necessity of a hate speech dataset fitted to the language that best reflects the culture. Trained with our dataset, we report the baseline performance of the models built on top of large pretrained language models. We also show that title information serves as context and is helpful to discern the target of hatred, especially when they are omitted in the comment.

show abstract

“…Recent works (Kedia and Nandy, 2021;Sharif et al, 2021;Jayanthi and Gupta, 2021) have explored various transformer-based models and some (Saha et al, 2021;Zhao and Tao, 2021) have made an ensemble of different ones which are focused on classification task. Offensive Span identification is in its developing stage, (Pavlopoulos et al, 2021) was the first to introduced a shared task and Offensive Span dataset.…”

Section: Related Workmentioning

confidence: 99%

NITK-IT_NLP@TamilNLP-ACL2022: Transformer based model for Toxic Span Identification in Tamil

LekshmiAmmal¹,

Ravikiran²,

M³

2022

Proceedings of the Second Workshop on Speech and Language Technologies for Dravidian Languages

View full text Add to dashboard Cite

Offensive Span identification in Tamil is a shared task that focuses on identifying harmful content, contributing to offensiveness. In this work, we have built a model that can efficiently identify the span of text contributing to offensive content. We have used various transformer-based models to develop the system, out of which the fine-tuned MuRIL model was able to achieve the best overall character F1-score of 0.4489.

show abstract

SemEval-2021 Task 5: Toxic Spans Detection

Cited by 49 publications

References 27 publications

Vietnamese Hate and Offensive Detection using PhoBERT-CNN and Social Media Streaming Data

Vietnamese Hate and Offensive Detection using PhoBERT-CNN and Social Media Streaming Data

KOLD: Korean Offensive Language Dataset

NITK-IT_NLP@TamilNLP-ACL2022: Transformer based model for Toxic Span Identification in Tamil

Contact Info

Product

Resources

About