Protocol for the automatic extraction of epidemiological information via a pre-trained language model

Area of Science:

• Computational epidemiology
• Natural Language Processing
• Public Health Informatics

Background:

• Timely extraction of epidemiological data from COVID-19 cases is crucial for effective public health interventions.
• Existing methods for data extraction are often manual and slow, hindering rapid response.
• Open-access COVID-19 case data requires efficient processing for public health surveillance.

Purpose of the Study:

• To present a protocol for a COVID-19 Cases Information Extraction (CCIE) system.
• To demonstrate the use of a pre-trained language model for automated data extraction.
• To outline steps for data preparation, model execution, and evaluation.

Main Methods:

• Development of a protocol for the CCIE system.
• Utilizing a pre-trained language model for Named Entity Recognition (NER) and text classification.
• Supervised training data preparation.
• Implementation of Python scripts for data processing.
• Machine evaluation and manual validation for assessing system performance.

Main Results:

• The CCIE system effectively extracts epidemiological fields from COVID-19 case data.
• The protocol details a reproducible method for automated information extraction.
• Both machine and manual validation confirm the system's effectiveness.

Conclusions:

• Automated extraction of epidemiological data using CCIE enhances the timeliness of public health measures.
• The presented protocol offers a robust solution for processing open-access COVID-19 case information.
• CCIE facilitates more efficient and responsive public health surveillance.