Application of LSTM emotion recognition based on deep learning and attention mechanism in patient relationship mediation

Medical disputes are a kind of disharmonious interpersonal relationship caused by different positions, views, and ways of thinking between doctors and patients. In recent years, it has shown an upward trend and has become a major issue affecting social stability and hospital development. This paper refers to the research process and results of Henkel et al. (2020) in the service industry and uses the CREMA-D dataset to train the model to obtain the accuracy of the algorithm and design an experimental process for using AI to solve doctor-patient relationships. This paper uses the CREMA-D dataset to divide emotions into six types: anger, disgust, fear, happy, neutral, and sad. Using natural language processing and LSTM to process emotions in speech data provides help and guidance for doctors to regulate patients' emotions, and helps doctors use AI integrated tools to regulate their own emotions and improve the happiness index of both parties.