A Large-Scale Dataset for Hate Speech Detection on Vietnamese Social Media Texts

Luu, Son T.; Nguyen, Kiet Van; Nguyen, Ngan Luu-Thuy

doi:10.1007/978-3-030-79457-6_35

Cited by 21 publications

(32 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On the other hand, we conducted a survey on the related works serving the task of classifying comments on social networks in Vietnamese, especially the Vietnamese Hate Speech Detection task is still modest [9,10,[19][20][21][22][23][24]. Specifically, the current studies revolve only based on two typical datasets by their outstanding high quality and large quantity of data points: ViHSD [10] and HSD-VLSP [9] datasets.…”

Section: Existing Hsd Modelsmentioning

confidence: 99%

“…We also overcome the remaining restrictions on the ViHSD and HSD-VLSP datasets in previous studies [9,10,[19][20][21][22][23][24] by proposing a two-phase data preprocessing techniques. Furthermore, we inherit the advantages of each study, such as the ability to conduct the experiment with deep learning, transfer learning models, and combined models.…”

Section: Hate Speech Detection With Streaming Datamentioning

confidence: 99%

“…We use two datasets: ViHSD [10] and HSD-VLSP [9] which contain 33,400 and 20,345 comments, respectively. Because the ViHSD and HSD-VLSP datasets are collected from social networking sites, they contain highly complex and diverse comments.…”

Section: Data Pre-processingmentioning

confidence: 99%

“…In order to do this, we used Word Segmenter of VnCoreNLP [13] for the PhoBERT model and NLTK [41] for the other models. Because the comments in both datasets, ViHSD [10] and HSD-VLSP [9], are raw text data, word segmentation is required to prepare the data for PhoBERT model training [36]. Moreover, PhoBERT used the VnCoreNLP RDRSegmenter [13] to preprocess the pre-training data [36] (including Vietnamese word and sentence segmentation), it is recommended that the same word segmenter must be used for PhoBERT downstream applications in relation to the input raw texts.…”

Section: Phasementioning

confidence: 99%

“…HSD and sentiment analysis have been inextricably linked [8], and these tasks have recently become popular topics in Natural Language Processing. In this section, we summarize the Vietnamese HSD task [9,10]. This task aims to detect whether a comment on social media is HATE, OFFENSIVE, or CLEAN.…”

Section: A Brief Introduction To Hate Speech Detection Taskmentioning

confidence: 99%

See 4 more Smart Citations

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