Computational Modeling of Science Communication with Natural Language Processing

Abstract

From the evolution of human beings to state-of-the-art artificial intelligence, knowledge created by scientists has been actively communicated to the public and has further driven the development of human society. Science communication is the process of conveying scientific information produced by the science community to the general public. Science communication not only aims to promote public understanding of scientific facts but also is tasked with encouraging public engagement with scientific information. For science communicators, effectively communicating science is challenging as it involves many key steps, including modeling the public's interests, identifying newsworthy information, and translating technical jargon into plain language that people can easily understand and engage with. My dissertation focuses on building natural language processing models and conducting large-scale analyses to understand how scientific information is communicated to the public. More specifically, in the three proposed studies, I developed NLP models that can automatically analyze the textual certainty in scientific claims, the holistic information change in science communications, and the public perception of science news stories. Further large-scale analysis demonstrates that the proposed methods could be used to analyze science communication at scale and provide valuable insights about effective science communication.