Enhancing Emotion Recognition in Text Data Based on Bi-LSTM and Attention Approach
Zhuojun Lyu · Advances in computer science research · 2024
Emotion recognition stands as a cornerstone across various domains, propelling the evolution of artificial intelligence.This paper introduces a pioneering approach to emotion recognition, employing a Bi-directional Long Short-Term Memory (Bi-LSTM) neural network fused with an attention mechanism (Att).The primary aim is to enrich the representation of emotional features within text data, particularly for sentiment analysis endeavors.The Bi-LSTM network effectively captures bidirectional dependencies within text sequences, while the Att meticulously focuses on pivotal segments of the input text, thereby enhancing performance and accuracy.Through experimentation on the Sentiment140 dataset, the model's efficacy is demonstrated, showcasing heightened accuracy and adaptability in contrast to conventional methods.The fusion of Bi-LSTM with Att presents a promising pathway for advancing sentiment analysis tasks, offering valuable insights into the intricacies of emotion recognition within text data.The outcomes of this research not only hold significance for social media analysis and intelligent customer service but also pave the way for potential applications in medical domains such as emotional health monitoring and mental illness diagnosis.Thus, fostering the application of artificial intelligence in diverse fields.