Crowd-Sourced Lexicons and Genetic Algorithms for Accurate Hate Speech Detection in Social Media
3rd International Conference on Machine Intelligence and Smart Innovation, ICMISI 2026, Alexandria, Mısır, 9 - 11 Mayıs 2026, ss.49-54, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/icmisi69868.2026.11584224
- Basıldığı Şehir: Alexandria
- Basıldığı Ülke: Mısır
- Sayfa Sayıları: ss.49-54
- Anahtar Kelimeler: Bag-of-Words, Dynamic Thresholding, Feature Extraction, Genetic Algorithm, Hashtags, Mentions, N-grams, Part-of-Speech Tagging, Readability Scores, Sentiment analysis, t-SNE, TFIDF
- Gazi Üniversitesi Adresli: Evet
Özet
Detecting hate speech becomes a persistent challenge because of the evolving slang and terminologies that requires adaptable techniques to these linguistic developments. To treat this challenge, a hybrid protocol is proposed that incorporate a sophisticated feather extraction approaches with applying Dynamic Thresholding on Genetic Algorithms (DTGAs) to emphasize accurate compilation. A diversity of powerful methodologies are accomplished for feature extraction approaches, such as Sentiment Analysis, Bag-of-Words (BoW), and Term Frequency-Inverse Document Frequency (TF-IDF), to expose substantial manner from textual data. For excavating recurring grammatical cues from the abusive content, additional methodologies are employed such as N -grams and Part-of-Speech (POS). To distinguish between direct and aggressive tone readability metrics are also utilized in addition to mentions and hashtags which facilitate definite targets precisely. Furthermore, another challenge appears that associated with considerable amount of data, thereby -SNE (tdistributed Stochastic Neighbor Embedding) is employed to facilitate the feather space and permit access to hidden pattern efficiently. For achieving the optimal performance, proposed DTGA adapts classification thresholds at the system's core and fine-tunes feature selection. Although, the diversity in digital environments, this proposed DTGA is an effective solution for real-time application, as experimental evaluations confirm its capability to discover hate speech more efficiently.