Roman Paolucci

Roman Paolucci

MS Student at Columbia University

Sitemap

A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.

Pages

Posts

portfolio

publications

How Many Words is a Picture Worth? Using Emojis from Social Media to Predict Future Stock Returns

Published in Working Paper, 2023

Using a new and comprehensive sample of more than 87 million Twitter posts referencing Russell 3000 firms between 2012 and 2022, we introduce a novel, unsupervised method of scoring the sentiment of emojis. Our method generates point-in-time dictionaries that map individual emojis to the contextual sentiment of recent tweets that contain them. In out-of-sample tests, we find that even controlling for the sentiment extracted from words, news, and corporate events, emoji sentiment correctly predicts future firm-level stock returns. Importantly, we show a newly emergent generation of Twitter users drive emoji-based return predictability, while more experienced users better predict returns using words. Understanding the sentiment of emojis has become increasingly important as individuals and market professionals continue to adopt these new forms of communication.

Recommended citation: Fox, Corbin, Eric K. Kelley, and Roman Paolucci. "How Many Words is a Picture Worth? Using Emojis from Social Media to Predict Future Stock Returns." Using Emojis from Social Media to Predict Future Stock Returns (March 15, 2023) (2023).
Download Paper

talks

Jumping from Volatility Surface to Option Price

Published:

Mitigating the model calibration issue by directly learning the map between the model parameter set and volatility surface to the option price trained on synthetic volatility surfaces generated by variational autoencoders.

teaching

Artificial Intelligence

High School Honors Course, Saddle River Day School, Department of Computer Science, 2023

Course Description: This course provides an in-depth exploration of advanced AI topics, including Machine Learning, Computer Vision, Generative Structures, Reinforcement Learning, Transformer Models, and Large Language Models (LLMs). Students will gain hands-on experience with cutting-edge techniques and tools, preparing them for the rapidly evolving field of AI.

Computer Science 1

High School CP Course, Saddle River Day School, Department of Computer Science, 2023

*Course Description: This course is a gentle technical introduction to the world of computer science and its many areas of study. This course takes a severely practical approach to education deeply rooted in theory. Students will waste little time on abstractions and dive right into learning to code as a tool for solving meaningful problems. Python is overwhelmingly the language of choice for academics and industrial practitioners alike and is where the course will begin. *

Data Science

High School Honors Course, Saddle River Day School, Department of Computer Science, 2023

Course Description: This course offers a comprehensive study of Data Science, encompassing application development, probability theory, and machine learning. Students will learn to build data-driven applications, apply probabilistic models, and utilize machine learning algorithms to extract insights and make data-driven decisions.

Natural Language Processing

High School Honors Course, Saddle River Day School, Department of Computer Science, 2023

Course Description: Natural language processing (NLP) is arguably the most important subfield of linguistics essentially teaching mathematical models to interpret, understand, extrapolate, etc. natural language. This course introduces natural language processing in the context of practical applications from sentiment analysis to text generation all of which hold a variety of applications from stock trading signal development to AI chat bots.