A New Approach for Sentiment Analysis of Sexual Harassment in Thai Sentences Using Transformer Models

A major issue that can occur in a number of contexts, including public areas and workplaces, is sexual harassment. One type of sexual harassment that can affect someone's mental health and interfere with everyday living is verbal harassment. This kind of harassment includes using sexually suggestive language, disparaging remarks, or statements that unnerve others. The victim is put in an intimidating or hostile setting, which can cause worries, worry, and mental health issues. The prevalence of verbal sexual harassment has surged recently, mostly as a result of the expansion of online communities that provide rapid messaging or commenting among members. For academics, analyzing words that can be indicative of verbal sexual harassment is a difficult undertaking. This work aims to identify sexual harassment in Thai literature and categorize the emotion using deep learning methods from the social networking site X (Twitter). The RoBERTa model was the main focus of this investigation, and it was contrasted with the BERT, mBERT, and BERT-th models. The percentages of accuracy were 91%, 95%, and 95%, in that order. RoBERTa outperformed the other two models in our analysis, with the greatest accuracy of 91%.