A comparison of automatic Boolean query formulation for systematic reviews
- PDF / 2,443,504 Bytes
- 26 Pages / 439.37 x 666.142 pts Page_size
- 68 Downloads / 195 Views
A comparison of automatic Boolean query formulation for systematic reviews Harrisen Scells1,2 · Guido Zuccon1 · Bevan Koopman2 Received: 1 July 2020 / Accepted: 6 October 2020 © Springer Nature B.V. 2020
Abstract Systematic reviews are comprehensive literature reviews that target a highly focused research question. In the medical domain, complex Boolean queries are used to identify studies. To ensure comprehensiveness, all studies retrieved are screened for inclusion or exclusion in the review. Developing Boolean queries for this task requires the expertise of trained information specialists. However, even for these expert searchers, query formulation can be difficult and lengthy: especially when dealing with areas of medicine that they may not be knowledgeable about. To this end, two computational adaptations of methods information specialists use to formulate Boolean queries have been proposed in prior work. These adaptations can be used to assist information specialists by providing a good starting point for query development. However, a number of limitations with these computational methods have been raised, and a comparison between them has not been made. In this study, we address the limitations of previous work and evaluate the two. We found that, between the two computational adaptions, the objective method is more effective than the conceptual method for query formulation alone, however, the conceptual method provides a better starting point for manual query refinement. This work helps to inform those building search tools that assist with systematic review construction. Keywords Information retrieval · Systematic reviews · Boolean queries · Query formulation
* Harrisen Scells [email protected] Guido Zuccon [email protected] Bevan Koopman [email protected] 1
The University of Queensland, Brisbane, QLD, Australia
2
CSIRO, Brisbane, QLD, Australia
13
Vol.:(0123456789)
Information Retrieval Journal
1 Introduction Systematic reviews are highly important within the medical domain. They are used to inform clinical decision making, and are seen as the highest form of medical evidence (Lavis et al. 2005). The process for developing a systematic review has many steps, and requires the support of clinical researchers, librarians, and review committees (McGowan and Sampson 2005). Systematic reviews are guided by a highly specific research question, and executed though a methodological study protocol (Chandler et al. 2019). Figure 1 illustrates the particular phase in systematic review creation that this study targets. Arguably, one of the most important processes in the creation of a systematic review is the identification of medical literature which will be synthesised later in the process. This identification process involves searching and screening studies (e.g., randomised controlled trials) from large medical databases (e.g., PubMed, which contains approximately 30 million studies at the time of writing). Screening literature is an important task that constitute a significant amount of
Data Loading...
