Code repository link: https://github.com/daxus4/emoji_impact
The version used in the paper is 515efc7
This project investigates the impact of emojis and sentiment-bearing words on financial sentiment classification in social media. Using over 1.4 million StockTwits messages and a RoBERTa-based model, controlled ablation experiments are conducted to measure the contribution of each component to bullish/bearish classification. Results show that sentiment-bearing words provide the strongest predictive signal, while emojis contribute additional information, highlighting the importance of preprocessing choices in financial NLP applications.
Authors: Davide Raffaelli, Daniele Marazzina, Simone Formentin
Title: Impact of emoji and sentiment-bearing words on financial social-media sentiment classification
Journal: Finance Research Letters
Link: https://www.sciencedirect.com/science/article/pii/S1544612326015837
DOI: https://doi.org/10.1016/j.frl.2026.111057
@article{RAFFAELLI2026111057,
title = {Impact of emoji and sentiment-bearing words on financial social-media sentiment classification},
journal = {Finance Research Letters},
volume = {114},
pages = {111057},
year = {2026},
issn = {1544-6123},
doi = {https://doi.org/10.1016/j.frl.2026.111057},
url = {https://www.sciencedirect.com/science/article/pii/S1544612326015837},
author = {Davide Raffaelli and Daniele Marazzina and Simone Formentin},
}