CRII: CHS: Facilitating Consumption and Re-expression of Scientific Information in a Journalism Context

University Of Washington, Seattle WA

Investigators

Abstract

This research will result in software tools that help journalists, scientists, and news readers write and read about scientific findings in more understandable ways. Though it is important that non-scientific audiences are able understand new advances in technological, biological, genetic, and other scientific fields, much of science writing is difficult to understand due to the use of jargon. One way to make scientific findings easier to understand is to re-express them using simpler terms, such as using the term "butterfly" instead of "lepidtoperan." By writing software to analyze collections of scientific publications and news articles, we will discover more understandable expressions for scientific terms, and develop user interface tools that allow an author or a reader to access these terms while they write or read about science. In addition to supporting science communication through the news, we expect our tools will help scientists create and share more broadly understandable summaries of their work. This work will contribute new software-based techniques to facilitate lexical simplification of scientific terminology, which draws upon and contributes to the fields of natural language processing and human-computer interaction. The technical objective of this project is to investigate how to enable effective re-expressions of scientific terms in tools through: 1) identifying strategies that human authors like science journalists use to re-express complex scientific content, and 2) developing user interfaces and algorithms that present simplifications to an author like a journalist as she writes about science for a general audience, or to a reader who lacks domain expertise. To build the simplification algorithms that will be used in our interactive tools, we will first learn mappings between scientific and simpler terms from corpora that include scientific writing such as journal publications in PLOS or PubMed as well as simpler summaries of this information written by the scientists or written by news journalists. We will apply word embeddings and other word association methods to such corpora to determine the effectiveness of different methods for learning simplifications. We will apply these simplifications in mixed initiative authoring and reading interfaces that enhance the capabilities of a human author or reader by recommending and enabling easy application of simplifications in various ways.