Predicting Hate Speech Targets on Social Media
Author Information
Author(s): Sahrish Khan, Rabeeh Ayaz Abbasi, Muddassar Azam Sindhu, Sachi Arafat, Akmal Saeed Khattak, Ali Daud, Mubashar Mushtaq
Primary Institution: Quaid-i-Azam University, Islamabad, Pakistan
Hypothesis
Can we predict which users are likely to become targets of hate speech on microblogging platforms?
Conclusion
The study successfully developed a framework that predicts potential targets of hate speech, achieving an accuracy of 93% with the Naïve Bayes classifier.
Supporting Evidence
- The proposed Hate Speech Target Prediction Framework (HTPK) utilizes machine learning to identify potential targets of hate speech.
- The Hate Speech Target Dataset (HSTD) contains 1500 tweets labeled as targets and 1500 as non-targets.
- Naïve Bayes classifier achieved the highest accuracy of 93% in predicting hate speech targets.
- Feature extraction methods included TF-IDF, N-grams, and Part-of-Speech tagging.
- Advanced transformer models were also tested, achieving an ensemble accuracy of 89.29%.
Takeaway
This study helps us understand who might get bullied online by looking at their posts, so we can warn them before it happens.
Methodology
The study used a combination of TF-IDF, N-grams, and Part-of-Speech tagging to analyze tweets and predict hate speech targets using various machine learning algorithms.
Potential Biases
Potential bias due to the exclusion of verified users, who may have different interaction patterns.
Limitations
The study excluded verified users, which may limit the generalizability of the findings.
Participant Demographics
The study focused on common users of microblogging platforms, excluding verified accounts.
Statistical Information
P-Value
0.0001
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website