Automated Event Extraction for News-Based Counterdata by Prof. Katie Keith, Computer Science, Williams College, Statistics Colloquium, Wednesday, October 19, 1:10 – 1:50 pm, North Science Building 114, Wachenheim.
Abstract: In the absence of data from official sources, how can one create quantitative analyses or data-driven policies? This talk focuses on creating counterdata—grassroots collection of missing datasets—by automatically extracting events from collections of news articles. News-based counterdata uses natural language processing, a branch of computer science that makes human language accessible to computers, and converts unstructured text into structured databases. This talk will present annotated corpora, natural language processing models, and empirical evaluation for several automated counterdata event extraction tasks.