Model of Mental Health Intervention based on Emotion Recognition - using Genetic Algorithm

In order to address the urgent problems of inadequate assessment indices and imperfect intervention strategies in recent years, this paper presents an innovative method using RoBERTa (BERT Robust Optimization Method), which is improved by means of a genetic algorithm. First, the text is converted into a compact vector representation by means of a complex word-embedding model. Then, the emotional features embedded within the text were meticulously extracted using advanced natural language processing techniques. Then, an emotion classifier was built to analyse and calculate the frequency of emotional words present in the text. In order to improve the performance of the model, we used a Genetic Algorithm to find an optimum combination between the model parameters and the structure. This included tuning the hyperparameters and tuning the parameters of the emotion classifier. Our results have shown that RoBERTa has been shown to be able to perform well on emotional recognition tasks, which has been shown to be highly accurate and efficient.