Events
[CSSL@CUHK Webinar] Applying Transformers to Predict Life Course Sequences
29 Oct 2024
4:00 pm to 5:30 pm (UTC+8,HKT)
Webinar
Dr. Linda Vecgaile
Linda Vecgaile holds a PhD in Political and Social Sciences from the European University Institute, Italy. She is currently a Research Scientist at the Max Planck Institute for Demographic Research, Department of Digital and Computational Demography, in Germany. Her research focuses on adapting natural language processing (NLP) models to predict life course sequences, offering insights into individuals’ familial and career trajectories. Her work also explores the broader implications of these patterns for social inequalities and policy development.
This study builds on life course theory, focusing on predicting future life events (ages 56-60) based on past sequences (ages 18-55). Using the Transformer encoder-decoder framework, we treat life events as sequential data, similar to words in a sentence, to capture patterns and relationships over time. With only 11 social employment states and basic demographic information, we develop a simple yet generalizable model. Our transformer model achieves above 80% accuracy in predicting life transitions, particularly for individuals with stable life paths. It also identifies deviations from recent patterns, highlighting the impact of earlier life experiences on future outcomes.
For details: https://linktr.ee/cssl.cuhk