Presenter: Theodoros Diakonidis
Authors: Diakonidis, Theodoros
Session: Tutorials 9
Title: ScreenmedR A Package for Automatizing the screening of Publications for Systematic Reviews – Meta-Analyses
Abstract: What follows is the introduction of a new program in R computing language, which aims to save every researcher a lot of time often spent in sifting publications coming from a Pubmed search, in order to identify relevant information for a systematic review or a meta-analysis. Supplying an input of 4-5 publications, the researcher believes that belong to the study, the program and most specifically the function screenmed(), can diminish the initial pubmed search up to 80% with almost 100% accuracy. The program accesses the abstracts of the publications and applies text mining methods in combination with hierarchical clustering, to build groups of abstracts on grounds of text resemblances. As a second step, it checks the resemblances between the groups of abstracts and the relevant 4-5 publications using cosine similarity in order to sort out the suitable ones. It also provides 2 functions which use mesh terms definitions to recognize similarities between publications. mesh_clean_bq()to identify common mesh terms between two groups of publications and mesh_by_name_bq()to sort out specific mesh terms from a group of publications.
GitHub repository: [ Ссылка ]

Links: [ Ссылка ]
[ Ссылка ]

свернуть