Text as Data conference highlights computational social science research

The 10th annual Text as Data conference, sponsored by the Institute for Research in the Social Sciences and Stanford Department of Political Science, was a two-day event that featured social and computer scientists sharing innovative statistical methods for natural language processing across diverse text corpora. SmartNews and VMware also served as generous corporate sponsors. The topics ranged from applications of word embedding methods to novel data collection methods using multilingual Twitter data to explore the impact of authoritarian regimes and the complex behavior of online social movements. Highlighted topics included model validation, detection of algorithmic bias, and the challenges of causal inference using machine learning on complex, high-dimensional text data.