Opinion mining, also known as sentiment analysis, involves classifying subjective sentiments in text into three categories: positive, negative, and neutral. This process delves into words and phrases to uncover the emotions expressed within sentences, paragraphs, or entire documents. Effective sentiment analysis relies on thorough text preprocessing, including the removal of irrelevant characters, stop words, and punctuation, as well as tokenization, which breaks text into meaningful units while preserving contextual relationships. When applied to Twitter (X) data, sentiment analysis faces unique challenges due to the informal language, abbreviations, hashtags, and extraneous characters often present in tweets. Comprehensive preprocessing is essential to address these issues, enabling the extraction of meaningful insights. Among the various methods available, Long Short-Term Memory (LSTM) networks excel in sentiment analysis because of their ability to capture sequential and contextual nuances inherent in textual data. This work proposes leveraging an LSTM-based model to perform opinion mining on Twitter (X) data related to the Indian election. The proposed model demonstrates exceptional performance, achieving an F1 score of 97.47% and an accuracy rate of 97.78 %. These results highlight not only the robustness of the LSTM approach but also its superiority in outperforming competing models for sentiment analysis tasks.

Sentiment analysis Opinion mining LSTM Twitter data Indian elections machine learning public opinion opinion mining