ISSN 2360-7998
Abstract
This study investigates the role of adverbial cues in emotion recognition across social media platforms using a machine learning approach. While most sentiment analysis tools focus on adjectives and emoticons, this research identifies adverbials as underexplored but powerful indicators of affective meaning. Using a dataset of 80,000 posts collected from Twitter, Facebook, and Instagram between 2022 and 2024, we annotated and classified emotional adverbials based on Appraisal Theory’s subsystems of Affect, Judgement, and Appreciation. Feature extraction techniques were applied using part-of-speech tagging, dependency parsing, and semantic vectorisation. Several machine learning algorithms—including Support Vector Machines (SVM), Random Forests, and Bidirectional LSTM—were trained and evaluated for performance. The study finds that models incorporating adverbial features outperform baseline models in emotion detection accuracy. The work bridges theoretical linguistics and applied NLP, offering insights for computational sentiment analysis, forensic linguistics, and digital communication.
Keywords: Adverbial, Machine learning, Emotion detection, Facebook, Instagram