New approach improves systematic reviews of scientific literature

Catherine Blake
Catherine Blake, Professor and Associate Dean for Academic Affairs

Risk assessments are conducted to determine if a chemical found in the environment is harmful to public health; for example, answering questions such as "does chemical 'x' promote cancer?" Conducting an impartial analysis of chemicals is thus critical to ensure that public policies reflect the best available scientific evidence. Unfortunately, the process of retrieving, extracting, and analyzing findings reported in scientific literature is time consuming and can delay when policies are updated to reflect new evidence.

Professor Catherine Blake and Jodi A. Flaws, professor of comparative biosciences at the University of Illinois, have developed an automated approach that moves beyond the retrieval of relevant literature to the extraction step of the information synthesis process. In a recent study of cell death and proliferation—two fundamental hallmarks of cancer—they demonstrate how simply counting the number of outcomes shows a very different picture than focusing on how key outcomes have changed.

"Systems currently just focus on the retrieval step, and if you base decisions solely on the number of abstracts retrieved, you would make the wrong decision," said Blake. "You have to look at the directionality of the evidence."

The natural language processing (NLP) system that Blake developed scales to over 400,000 abstracts and identifies the directionality of evidence (refuting, neutral, or supporting) for 27 different chemicals. Their approach automates the extraction step, providing researchers with waffle plots that visually present the distribution of supporting, neutral, and refuting evidence for each chemical. For example, in the waffle plots below, chemical 1 has more refuting evidence whereas chemical 22 has more supporting evidence.

waffle plots of 27 chemicals showing refuting and supporting evidence

This automated approach provides researchers with important detail that is missing in existing automated systems, which is closer to the manual processes used in decision-making, and also maintains the level of transparency needed in a public policy setting. Blake and Flaws' study, "Using semantics to scale up evidence-based chemical risk-assessments," was recently published in the peer-reviewed open access journal PLOS One.

Blake's research seeks to accelerate science and inform policy by automatically extracting and summarizing claims reported in the scientific literature. She holds a PhD and MS in information and computer science from the University of California, Irvine, and an MS and BS in computer science from the University of Wollongong, Australia.

Updated on
Backto the news archive

Related News

Faculty receive support for AI-related projects from new pilot program

Associate Professor Yun Huang, Assistant Professor Jiaqi Ma, and Assistant Professor Haohan Wang have received computing resources from the National Artificial Intelligence Research Resource (NAIRR), a two-year pilot program led by the National Science Foundation in partnership with other federal agencies and nongovernmental partners. The goal of the pilot is to support AI-related research with particular emphasis on societal challenges. Last month, awardees presented their research at the NAIRR Pilot Annual Meeting.

iSchool participation in iConference 2025

The following iSchool faculty and students will participate in iConference 2025, which will be held virtually from March 11-14 and physically from March 18-22 in Bloomington, Indiana. The theme of this year's conference is "Living in an AI-gorithmic world."

Carboni joins the iSchool faculty

The iSchool is pleased to announce that Nicola Carboni has joined the faculty as an assistant professor. He previously served as a postdoctoral researcher and lecturer in digital humanities at the University of Geneva.

Nicola Carboni

Youth-AI-Safety named a winning team in international hackathon

A team of researchers from the SALT (Social Computing Systems) Lab has been selected as a winner in an international hackathon hosted by the Berkeley Center for Responsible, Decentralized Intelligence. The LLM Agents MOOC Hackathon brought together over 3,000 students, researchers, and practitioners from 127 countries to build and showcase innovative work in large language model (LLM) agents, grow the AI agent community, and advance LLM agent technology.

Chan to present "Predatory Data" work at named lectures

Associate Professor Anita Say Chan will present research drawn from her new book, Predatory Data: Eugenics in Big Tech and Our Fight for an Independent Future, at two named lectures this month. The lectures, which celebrate Women's History Month, will be held at the University of Minnesota and Carnegie Mellon University.

Anita Say Chan