cv

Principal Researcher

socius

socius is an independent research lab endorsed by the London School of Economics and Political Science. The lab is rooted in social science principles, believing that AI development and deployment must be guided by robust theories and philosophies. We create tools and frameworks that both technical and social science researchers can confidently use and interpret.

• NLP

Junior NLP Engineer

Numen Capital

Worked on the in-house Knowsis research team, leveraging NLP and machine learning for alternative data insights. (1) Developed NLP pipelines for alternative data insights, including ESG tweet classifier based on transformer architecture (97% accuracy), rule-based sentiment analyser, and domain-specific dictionary. (2) Analysed alternative data signals derived from NLP pipelines using time series models. This analysis led to accurate forecasts of market movements, directly assisting quantitative investment strategies.

• NLP

MSc

University of London

Data Science and Artificial Intelligence

• Big data analysis
• Data programming in Python
• Statistics and statistical mining
• Machine learning
• Data science research topics
• Mathematics of financial markets
• Artificial intelligence
• Natural language processing
• Neural networks
• Social media and network science

BSc

London School of Economics and Political Science

Politics and Economics

• Macroeconomic Principles
• Microeconomic Principles I
• Public Choice and Politics
• Research Design in Political Science
• Monetary Economics
• Applied Quantitative Methods for Political Science
• Algorithms and Programming
• Game Theory I
• Politics of Money and Finance in Comparative Perspective

sentibank: A Unified Resource of Sentiment Lexicons and Dictionaries

International AAAI Conference on Web and Social Media

Sentiment analysis is critical across computational social science domains, but faces challenges in interpretability. Rule-based methods relying on expert lexicons enable transparency, yet applying them is hindered by resource fragmentation and lack of validation. This paper introduces sentibank, a large-scale unified database consolidating 15 original sentiment dictionaries and 43 preprocessed dictionaries, spanning 7 genres and 6 domains.

RedditHarbor

Open-source Python toolkit designed to simplify the process of collecting and archiving Reddit data for research purposes

• Reddit Data Collection
• Research Data Toolkit
• Research Data Pipeline
• Reddit Data API
• Data Crawler