A novel application of machine learning and zero-shot classification methods for automated abstract screening in systematic reviews - read online

Aug 5, 2026

Upload by Jiven

Zero-shot classification assigns labels to text without prior training by learning how to encode a question and retrieve its answer from the text. Systematic reviews in evidence-based domains like health sciences require extensive expert effort in collecting, filtering, evaluating, and synthesising large literature sets, with abstract screening being especially time-consuming and subjective. The paper presents a novel use of traditional machine learning and zero-shot methods for automated abstract screening, reporting competitive accuracy, precision, and recall across seven public datasets and suggesting reduced screening burden and human error.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Introduction

Steps of systematic reviews

Motivation for automation with ML and NLP

Focus of the proposed approach

A novel application of machine learning and zeroshot classification methods for automated abstract screening in systematic reviews . MORENO-GARCIA, C.F., JAYNE, C., ELYAN, E. and ACEVES-MARTINS, M. 2023 © 2023 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY license ([http://creativecommons.org/licenses/by/4.0](http://creativecommons.org/licenses/by/4.0/)[/](http://creativecommons.org/licenses/by/4.0/)). A novel application of machine learning and zero-shot classification methods for automated abstract screening in systematic reviews Carlos Francisco Moreno-Garcia a,∗, Chrisina Jayne b, Eyad Elyan a, Magaly Aceves-Martins c a School of Computing, Robert Gordon University, Garthdee Road, Aberdeen, AB10 7QB, Scotland, UK b Teesside University, Southfield Road, Middlesbrough, TS1 3BX, England, UK c The Rowett Institute, University of Aberdeen, Ashgrove Road West, Aberdeen, AB25 2ZD, Scotland, UK A R T I C L E I N F O Keywords: Machine learning Systematic review Abstract screening Class imbalance Zero-shot classification A B S T R A C T Zero-shot classification refers to assigning a label to a text (sentence, paragraph, whole paper) without prior training. This is possible by teaching the system how to codify a question and find its answer in the text. In many domains, especially health sciences, systematic reviews are evidence-based syntheses of information related to a specific topic. Producing them is demanding and time-consuming in terms of collecting, filtering, evaluating and synthesising large volumes of literature, which require significant effort performed by experts. One of its most demanding steps is abstract screening, which requires scientists to sift through various abstracts of relevant papers and include or exclude papers based on pre-established criteria. This process is timeconsuming and subjective and requires a consensus between scientists, which may not always be possible. With the recent advances in machine learning and deep learning research, especially in natural language processing, it becomes possible to automate or semi-automate this task. This paper proposes a novel application of traditional machine learning and zero-shot classification methods for automated abstract screening for systematic reviews. Extensive experiments were carried out using seven public datasets. Competitive results were obtained in terms of accuracy, precision and recall across all datasets, which indicate that the burden and the human mistake in the abstract screening process might be reduced. 1. Introduction Review articles are common and crucial sources of knowledge across different domains. They provide a comprehensive study of an area ([e.g. health-related](e.g. health-related) topics, medical interventions, social sciences, etc.) and serve as a rich source of information for researchers in the respective field. In health sciences, but also in other domains such as social sciences [1], a systematic review (SR) approach is often followed to construct such comprehensive studies aiming to collect, summarise, critically evaluate, and synthesise knowledge. This practice is crucial to keeping clinicians and medical experts informed of the latest development in the field. The type of SR will depend on the research question and the type of data available to answer such questions. Based on the data, there are different approaches to conducting SRs [2]. Typical examples include the work presented by Aceves-Martins et al. [3], where the authors consulted different databases for articles exploring the relationship between obesity and oral health in Mexican children. Most SR approaches consist of the following steps [4]: 1. identify relevant databases of published peer-reviewed literature 2. use specific keywords and Boolean connectors to search for potentially relevant papers 3. screen titles and abstracts retrieved from the searches [4. read](4. read) full-tex

FAQs

What is zero-shot classification in the context of this work?

Zero-shot classification labels text without prior training by encoding a question and locating its answer in the text.

Why is abstract screening in systematic reviews difficult?

Abstract screening is time-consuming and subjective, requiring expert consensus that may not always be achievable.

How does the proposed method support automated abstract screening?

The paper applies traditional machine learning and zero-shot classification to automate abstract screening for systematic reviews, using seven public datasets.

Aug 5, 2026

Upload by Jiven