Semi-Supervised Self-Learning for Arabic Hate Speech Detection

Alsafari, Safa; Sadaoui, Samira

doi:10.1109/smc52423.2021.9659134

Cited by 11 publications

(8 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“… Gulf Arabic: This type is being spoken throughout the Gulf Cooperation Council (GCC), which includes Saudi Arabia, the United Arab Emirates, Qatar, Bahrain, Kuwait, and Oman. Its distinctive vocabulary and pronunciation serve as defining characteristics ( Alsafari & Sadaoui, 2021a ). …”

Section: Review Findings and Discussionmentioning

confidence: 99%

A systematic literature review of hate speech identification on Arabic Twitter data: research challenges and future directions

Alhazmi,

Mahmud,

Idris

et al. 2024

PeerJ Computer Science

View full text Add to dashboard Cite

The automatic speech identification in Arabic tweets has generated substantial attention among academics in the fields of text mining and natural language processing (NLP). The quantity of studies done on this subject has experienced significant growth. This study aims to provide an overview of this field by conducting a systematic review of literature that focuses on automatic hate speech identification, particularly in the Arabic language. The goal is to examine the research trends in Arabic hate speech identification and offer guidance to researchers by highlighting the most significant studies published between 2018 and 2023. This systematic study addresses five specific research questions concerning the types of the Arabic language used, hate speech categories, classification techniques, feature engineering techniques, performance metrics, validation methods, existing challenges faced by researchers, and potential future research directions. Through a comprehensive search across nine academic databases, 24 studies that met the predefined inclusion criteria and quality assessment were identified. The review findings revealed the existence of many Arabic linguistic varieties used in hate speech on Twitter, with modern standard Arabic (MSA) being the most prominent. In identification techniques, machine learning categories are the most used technique for Arabic hate speech identification. The result also shows different feature engineering techniques used and indicates that N-gram and CBOW are the most used techniques. F1-score, precision, recall, and accuracy were also identified as the most used performance metric. The review also shows that the most used validation method is the train/test split method. Therefore, the findings of this study can serve as valuable guidance for researchers in enhancing the efficacy of their models in future investigations. Besides, algorithm development, policy rule regulation, community management, and legal and ethical consideration are other real-world applications that can be reaped from this research.

show abstract

Section: Review Findings and Discussionmentioning

confidence: 99%

A systematic literature review of hate speech identification on Arabic Twitter data: research challenges and future directions

Alhazmi,

Mahmud,

Idris

et al. 2024

PeerJ Computer Science

View full text Add to dashboard Cite

show abstract

“…Hate or Not hate is the most common binary label used in literature, such as Alshaalan and Al-Khalifa ( 2020 ), Aref et al ( 2020 ), Omar et al ( 2020 ), Romim et al ( 2021 ), Saeed et al ( 2022 ), and Khezzar et al ( 2023 ). Others used different labels for binary classification, such as clean or offensive (Alsafari et al, 2020b ; Alsafari and Sadaoui, 2021 ), hateful or normal (Salomon et al, 2022 ). Some researchers were more precise in identifying the labels according to the type of hate speech detected.…”

Section: Literature Reviewmentioning

confidence: 99%

“…This helps researchers when scraping social media, to search for keywords that are more related to this dialect. Levantine (Duwairi et al, 2021 ) and Gulf (Alsafari et al, 2020b ; Alsafari and Sadaoui, 2021 ) are examples of dialects used by people in a wide region of the Arab world. So, when a researcher needs to collect data in the Levantine dialect, for example, they should add to their query the desired locations, including Jordan, Palestine, Syria, and Lebanon.…”

Section: Literature Reviewmentioning

confidence: 99%

Hate speech detection in the Arabic language: corpus design, construction, and evaluation

Ahmad,

Azzeh,

Alnagi

et al. 2024

Front. Artif. Intell.

View full text Add to dashboard Cite

Hate Speech Detection in Arabic presents a multifaceted challenge due to the broad and diverse linguistic terrain. With its multiple dialects and rich cultural subtleties, Arabic requires particular measures to address hate speech online successfully. To address this issue, academics and developers have used natural language processing (NLP) methods and machine learning algorithms adapted to the complexities of Arabic text. However, many proposed methods were hampered by a lack of a comprehensive dataset/corpus of Arabic hate speech. In this research, we propose a novel multi-class public Arabic dataset comprised of 403,688 annotated tweets categorized as extremely positive, positive, neutral, or negative based on the presence of hate speech. Using our developed dataset, we additionally characterize the performance of multiple machine learning models for Hate speech identification in Arabic Jordanian dialect tweets. Specifically, the Word2Vec, TF-IDF, and AraBert text representation models have been applied to produce word vectors. With the help of these models, we can provide classification models with vectors representing text. After that, seven machine learning classifiers have been evaluated: Support Vector Machine (SVM), Logistic Regression (LR), Naive Bays (NB), Random Forest (RF), AdaBoost (Ada), XGBoost (XGB), and CatBoost (CatB). In light of this, the experimental evaluation revealed that, in this challenging and unstructured setting, our gathered and annotated datasets were rather efficient and generated encouraging assessment outcomes. This will enable academics to delve further into this crucial field of study.

show abstract

“…Hate or Not hate is the most common binary label used in literature, such as Alshaalan and Al-Khalifa (2020), Aref et al (2020), Omar et al (2020), Romim et al (2021), Saeed et al (2022), and Khezzar et al (2023). Others used different labels for binary classi cation, such as clean or offensive (Alsafari et al, 2020b;Alsafari and Sadaoui, 2021), hateful or normal (Salomon et al, 2022). Some researchers were more precise in identifying the labels according to the type of hate speech detected.…”

Section: Hate Speech and Related Conceptsmentioning

confidence: 99%

Hate Speech Detection in the Arabic Language: Corpus Design, Construction and Evaluation

Ahmad,

Azzeh,

Alnagi

et al. 2023

Preprint

View full text Add to dashboard Cite

Hate Speech Detection in Arabic presents a multifaceted challenge due to the broad and diverse linguistic terrain. With its multiple dialects and rich cultural subtleties, Arabic requires particular measures to address hate speech online successfully. To address this issue, academics and developers have used natural language processing (NLP) methods and machine learning algorithms adapted to the complexities of Arabic text. However, many proposed methods were hampered by a lack of a comprehensive dataset/corpus of Arabic hate speech. In this research, we propose a novel multi-class public Arabic dataset comprised of 403,688 annotated tweets categorized as extremely positive, positive, neutral, or negative based on the presence of hate speech. Using our developed dataset, we additionally characterize the performance of multiple machine learning models for Hate speech identification in Arabic Jordanian dialect tweets. Specifically, the Word2Vec, TF-IDF, and AraBert text representation models have been applied to produce word vectors. With the help of these models, we can provide classification models with vectors representing text. After that, seven Machine learning classifiers have been evaluated: Support Vector Machine (SVM), Logistic Regression (LR), Naive Bays (NB), Random Forest (RF), AdaBoost (Ada), XGBoost (XGB), and CatBoost (CatB). In light of this, the experimental evaluation revealed that, in this challenging and unstructured setting, our gathered and annotated datasets were rather efficient and generated encouraging assessment outcomes. This will enable academics to delve further into this crucial field of study.

show abstract

Semi-Supervised Self-Learning for Arabic Hate Speech Detection

Cited by 11 publications

References 23 publications

A systematic literature review of hate speech identification on Arabic Twitter data: research challenges and future directions

A systematic literature review of hate speech identification on Arabic Twitter data: research challenges and future directions

Hate speech detection in the Arabic language: corpus design, construction, and evaluation

Hate Speech Detection in the Arabic Language: Corpus Design, Construction and Evaluation

Contact Info

Product

Resources

About