Screening evidence for systematic reviews using a text-mining system: The RobotAnalyst

Article type
Authors
Nolan K1, Ananiadou S2, Le Pogam M3, Von Elm E4, Przybyła P5, Mcleod C1
1NICE
2School of Computer Science, University of Manchester
3Cochrane Switzerland, Institute of Social and Preventive Medicine, Lausanne University Hospital, Lausanne University
4Cochrane Switzerland
5National Centre for Text Mining, University of Manchester
Abstract
Objectives:
•Describe experiences and difficulties of screening evidence for populating systematic reviews for public-health topics.
•Provide an overview of text-mining methods and demonstrate a system developed by the National Centre for Text Mining to support screening: the RobotAnalyst.
•Present results of an evaluation of RobotAnalyst for a number of systematic reviews.
•Explore the wider potential benefits of Robot Analyst for supporting evidence synthesis for guideline development.

Description: Text-mining methods have the potential to reduce time and costs for the development of evidence reviews.
The workshop will be structured into four parts:
1) A general discussion of the challenges of searching and screening evidence for public health systematic reviews.
2) An overview of text-mining methods and a demonstration of the Robot Analyst. This is a bespoke application developed to support the evidence-review process for the development of public-health guidelines.
3) Presentation of results from NICE and Cochrane Switzerland on the evaluation of the system. Results will be presented on the potential time savings, specificity of the text-mining functionality and value of additional features in the system for the generation of evidence reviews.
4) An interactive discussion on the transferability of text-mining functionality beyond public health including a international panel of experts who will take questions and comments from participants.