A LIGHTWEIGHT LSTM FRAMEWORK FOR CONTEXTUAL SENTIMENT CLASSIFICATION

With the rapid increase in opinion-rich content shared across the internet, text sentiment analysis has emerged as a vital tool in both academic research and industrial applications. Sentiment analysis typically involves classifying a piece of text as expressing positive, negative, or neutral emotion. Traditional approaches to text classification often require extensive feature engineering and rely heavily on tokenization and embedding techniques, making them resource-intensive and less adaptive to context. To address these limitations, Long Short-Term Memory (LSTM) networks—an advanced form of Recurrent Neural Networks (RNNs)—have been adopted for their ability to capture long-range dependencies in textual data. This study proposes a sentiment classification model based solely on LSTM architecture to analyze short texts and effectively extract context-aware sentiment patterns. Unlike conventional models, LSTM-based frameworks can learn temporal word relationships without explicit syntactic parsing or handcrafted features. By leveraging the memory capabilities of LSTM, the proposed model enhances sentiment categorization accuracy while maintaining a relatively lightweight computational profile. Experimental evaluations demonstrate the effectiveness of LSTM in capturing contextual semantics, making it a suitable choice for real-time sentiment detection tasks in dynamic and user-generated content environments.