Sentiment Analysis of Trending Topics on Social Media X Using Natural Language Processing and LSTM

Abstract

In today’s fast-paced digital era, trending news on Social Media X spreads rapidly, influences public opinion, and is often vulnerable to disinformation. This study analyzes netizens’ sentiment towards trending topics on Social Media X using Natural Language Processing (NLP) and a Long Short-Term Memory (LSTM) model. A dataset of 4483 comments was collected across 15 trending topics (Feb–Jun 2025). The preprocessing steps included cleansing, case folding, stopword removal, tokenization, and translation to handle bilingual data. Results show sentiment distribution: 35% positive, 36% negative, and 29% neutral. Model performance varied between 34%–67% accuracy, with precision, recall, and F1-scores indicating that topic sensitivity, language diversity, and data imbalance strongly influenced outcomes. This research contributes to text analytics by providing a baseline model for real-time trending news sentiment analysis in Indonesia, particularly under multilingual and noisy data conditions.In today’s fast-paced digital era, trending news on Social Media X spreads rapidly, influences public opinion, and is often vulnerable to disinformation. This study analyzes netizens’ sentiment towards trending topics on Social Media X using Natural Language Processing (NLP) and a Long Short-Term Memory (LSTM) model. A dataset of 4483 comments was collected across 15 trending topics (Feb–Jun 2025). The preprocessing steps included cleansing, case folding, stopword removal, tokenization, and translation to handle bilingual data. Results show sentiment distribution: 35% positive, 36% negative, and 29% neutral. Model performance varied between 34%–67% accuracy, with precision, recall, and F1-scores indicating that topic sensitivity, language diversity, and data imbalance strongly influenced outcomes. This research contributes to text analytics by providing a baseline model for real-time trending news sentiment analysis in Indonesia, particularly under multilingual and noisy data conditions.