Named Entity Recognition and Classification for Medical Prospectuses

Area of Science:

• Medical Informatics
• Natural Language Processing
• Computational Linguistics

Background:

• Medical knowledge extraction is crucial for developing clinical decision support systems.
• Accurate patient treatment relies on structured medical data.
• Minimizing medical errors necessitates efficient information processing.

Purpose of the Study:

• To develop and evaluate a method for extracting structured medical information from Romanian drug prospectuses.
• To identify key entities such as drug-treated conditions, medicine names, and drug types.
• To assess the accuracy of the information extraction process.

Main Methods:

• Utilized Stanford Named Entity Recognition (NER) Tagger.
• Trained the NER model on Romanian medical prospectuses.
• Tested the model with three distinct medication types.
• Calculated the accuracy of extracted data for each test case.

Main Results:

• Successfully extracted critical medical entities from prospectuses.
• Demonstrated the feasibility of using NLP for medical data.
• Quantified the accuracy of the developed extraction method.

Conclusions:

• The proposed method effectively extracts vital medical information from Romanian drug prospectuses.
• Extracted data can populate databases for decision-support applications.
• This approach aids in identifying optimal patient treatments and reducing errors.