gms | German Medical Science

Introduction: Search is the first stage in evidence synthesis, and the use of a valid and reproducible search strategy forms the foundation of the systematic review process. However, the key steps of developing, validating, visualising, saving, sharing, peer-reviewing, and translating search strategies (to the syntax of different databases) are often fragmented across a variety of unconnected, non-interoperable platforms. As a result, designing effective search strategies is often subject to errors and inefficiencies. We present an open-access platform that offers a unified approach to structured searching which promotes replicable methods and reproducible results.

Method: 2Dsearch is a radical alternative to conventional line-by-line query builders. Instead of entering Boolean strings into one-dimensional search boxes, queries are formulated by manipulating objects on a two-dimensional canvas. Query suggestions are provided via an NLP services API, and support is offered for optimising and translating search strategies for different databases. Moreover, strategies can be saved, shared and reviewed as executable artefacts. This approach eliminates many sources of error, makes the query semantics more transparent, and offers an open-access platform for sharing reproducible search templates and best practices.

Results: 2Dsearch currently supports over 1,200 registered users (and many more unregistered) in providing an open-access, integrated development environment specifically optimised for search strategy development. It includes:

• A visual framework which eliminates many errors associated with traditional command-line query formulation tools;
• Search results that update in real-time, and individual blocks with hit counts that can be enabled/disabled on demand;
• A platform-agnostic representation and support for multiple databases which mitigates inefficient ‘translation’ of search strategies across databases;
• Support for parsing existing text-based strategies and rendering them on the canvas as editable, executable objects, whilst also identifying common errors such as redundant structure, misplaced parentheses, and duplicate elements;
• Interactive query suggestions that avoid the problems of phrase boundary detection and ‘query drift’ that undermine traditional query expansion techniques;
• The ability for search strategies to be shared as executable objects, facilitating team working, collaboration and the effective reuse of templates and best practices.

Conclusions: Search is the first stage in evidence synthesis, and the use of a valid and reproducible search strategy forms the foundation of the systematic review process. We present an open-access platform that offers a unified approach to structured searching which promotes transparent methods and reproducible results.

Keywords: boolean, structured searching, living systematic reviews, automation, natural language processing