Text mining methods to support the development of sensitive search strategies in public health reviews

Tags: Workshop
Kontonatsios G1, Przybyła P1, Nolan K2, Shaw B2, Haynes C2
1National Centre for Text Mining, University of Manchester, 2National Institute for Health and Clinical Excellence (NICE), UK

Objectives: To cover text mining methods to support study identification in public health reviews. Specifically, we aim to:

1. discuss limitations of conventional keyword-based search engines (e.g. PubMed) that are ill-suited to the development of sensitive search strategies;

2. provide an overview of text mining methods for generating semantic metadata over large scale textual collections; and

3. demonstrate the use of semantically enriched search engines that enable interactive, exploratory searching of relevant evidence.

Description: The unstructured and ambiguous nature of natural language in public health literature, poses a barrier to the accessibility and discovery of information. We will discuss challenges to information discovery that are inherent in keyword-based search engines. We will then demonstrate potential solutions offered by semantic search systems, enhanced by text mining methods. Participants will receive an introduction to various semantic search features (e.g. faceted search, automatic query expansion, queries as natural language questions) and asked to construct complex queries using on-line semantic search systems. This will provide an appreciation of how text mining can support the development of sensitive search strategies. The intended outcome of this workshop is to highlight benefits and limitations of these emerging technologies.

Participants are encouraged to bring laptops to try the online search systems.