Predicting the victims of hate speech on microblogging platforms

Area of Science:

• Computational Linguistics
• Social Computing
• Machine Learning

Background:

• Hate speech is a significant issue on microblogging platforms.
• Existing research primarily focuses on content analysis for hate speech detection.
• There is a need to identify potential targets of hate speech proactively.

Purpose of the Study:

• To propose a novel Hate-speech Target Prediction Framework (HTPK).
• To introduce a new Hate Speech Target Dataset (HSTD) for target identification.
• To identify optimal features and compare machine learning algorithms for predicting hate speech targets.

Main Methods:

• Development of the Hate-speech Target Prediction Framework (HTPK).
• Creation of the Hate Speech Target Dataset (HSTD) with labeled targets and non-targets.
• Utilizing Term Frequency-Inverse Document Frequency (TFIDF), N-grams, and Part-of-Speech (PoS) tags as features.
• Evaluation of various machine learning algorithms, including Naïve Bayes (NB).

Main Results:

• The Naïve Bayes classifier achieved the highest accuracy at 93%.
• The proposed HTPK framework demonstrated effectiveness in predicting hate speech targets.
• TFIDF, N-grams, and PoS tags were identified as optimal features for target prediction.

Conclusions:

• The research provides a foundation for proactive hate speech mitigation strategies.
• The study highlights the effectiveness of the HTPK framework and the NB classifier.
• Identifying potential targets is crucial for developing more effective interventions against online hate speech.