Mining the neuroimaging literature

Automated analysis of the biomedical literature (literature mining) offers a rich source of insights. However, such analysis requires collecting a large number of articles and extracting and processing their content. This task is often prohibitively difficult and time-consuming. Here, we provide tools to easily collect, process, and annotate the biomedical literature. In particular, https://neuroquery.github.io/pubget/pubget.html is an efficient and reliable command-line tool for downloading articles in bulk from PubMed Central, extracting their contents and metadata into convenient formats, and extracting and analyzing information such as stereotactic brain coordinates. https://jeromedockes.github.io/labelbuddy/labelbuddy/current/ is a lightweight local application for annotating text, which facilitates the extraction of complex information or the creation of ground-truth labels to validate automated information extraction methods. Further, we describe repositories where researchers can share their analysis code and their manual annotations in a format that facilitates reuse. These resources can help streamline text mining and meta-science projects and make text mining of the biomedical literature more accessible, effective, and reproducible. We describe a typical workflow based on these tools and illustrate it with several example projects.