User Intent Prediction and Response in Human-Computer Interaction via BiLSTM

Abstract

The purpose of this study is to explore user intent prediction and automatic response systems in human-computer interaction based on deep learning. By analyzing the relationship between user behavior characteristics and intention prediction and combining the BiLSTM (Bidirectional long short-term memory network) model, an accurate method of user intention prediction is proposed. The experimental results show that compared with traditional machine learning models (such as SVM and random forest), the BiLSTM model has significant advantages in prediction accuracy and system response-ability, especially in understanding and responding to user intentions in complex scenarios. In addition, this study also analyzes the correlation between homepage data (such as user click module, visit frequency, page stay time, etc.) and user intent and reveals the impact of different page elements on user behavior. Future research can further optimize the adaptability of the model, incorporate more multimodal data (such as voice, video, etc.), and explore personalized and context-aware user intent prediction methods to promote the development of more intelligent and naturalized human-computer interaction systems. The research results provide a theoretical basis and technical support for improving the human-computer interaction experience and promoting intelligent service applications.