MNVLD: Violent Language Detection Model with Optimized Loss Function
Dongfang Li, Jing Liu, Li Tan, Ziliang Shang, Zikang Liu · 2023
Violent language on the Internet has become a serious problem on social platforms, posing a threat to users’ sense of security and a conducive communication environment. Most existing methods focus on creating violent language dictionaries and classifying texts using support vector machines (SVM) and naive Bayesian techniques. However, these methods face challenges when confronted with tens of thousands of text comments or commenters using obscure words. Additionally, the current network violence language detection technology performs poorly in multi-domain settings. This paper aims to enhance the MFND model in the context of rumor detection. We propose the MNVLD model, which incorporates an appropriate loss function based on the characteristics of the network violence language dataset, to effectively detect violent language across multiple domains.