Text mining and prioritisation for literature triage
SWIFT-Review (Sciome Workbench for Interactive computer-Facilitated Text-mining) is a free, interactive desktop workbench designed to assist with systematic literature reviews and prioritization.It helps users manage and prioritize large volumes of academic literature (titles and abstracts) by using statistical text mining and machine learning methods._x000D_ Users upload bibliographic data, typically from sources like PubMed, in specific formats (PubMed XML, EndNote XML, etc.)_x000D_ The application automatically tags documents with relevant metadata (e.g., MeSH terms) and provides a search engine for users to filter the literature based on keywords or tags._x000D_ Users manually screen an initial set of documents (a "training seed"). Based on these inclusions/exclusions, SWIFT-Review builds a model to estimate the relevance probability of the remaining unscreened documents._x000D_ Documents are then ranked according to their estimated relevance using latent Dirichlet allocation (LDA) topic modeling, allowing screeners to review the most pertinent articles first, which significantly reduces the manual screening burden.
