Zhuge Repeated Arrowbow

Keyword: Mechine Learning, Emotion Recognition
Online surveys and research are primarily conducted in the form of text-based questionnaires. Compared to real-life conversations between individuals, pure text-based surveys are unable to capture the tone variations and emotional information conveyed by users during their narratives. To address this, we developed a chatbot that incorporates voice emotion recognition. By utilizing BLSTM (Bidirectional Long Short-Term Memory) models to capture emotional changes in users’ voice waveforms and spectrograms, the chatbot can actively guide users to generate more content when their emotions are stronger. This enriches the information collected in the surveys.