CovidNLP

CovidNLP distills and synthesizes peer-reviewed COVID-19 literature using natural language processing to reveal direct and systemic public health implications.

Key Features:

• Data Source: Uses the World Health Organization COVID-19 Global Literature repository comprising more than 5,000 peer-reviewed research articles covering epidemiology, clinical features, diagnosis, treatment, social factors, and economic impacts.
• Summarization Technique: Employs an extractive summarizer to condense articles into concise summaries while retaining salient information.
• Feature Space Exploration: Applies word embeddings to explore the feature space of summarized data and to visualize and analyze complex associations in the literature.
• Systemic Implications Analysis: Identifies and highlights systemic implications reported in the literature, including potential increases in tuberculosis (TB) and cancer mortality related to disruptions such as drug export lockdowns.
• Continuous Model Updates: Updates models with newly published literature to incorporate current peer-reviewed evidence.

Scientific Applications:

• Rapid literature synthesis: Condenses large volumes of COVID-19 research to support clinical management and research prioritization.
• Public health impact assessment: Synthesizes direct and indirect effects of the pandemic to inform decision-making and pre-emptive public health actions.
• Association visualization: Visualizes complex data associations from the literature to enhance understanding and facilitate strategic planning in public health responses.

Methodology:

Natural language processing using extractive summarization and word embeddings to summarize articles and explore feature spaces of the COVID-19 literature.