Development of Structured Data from Unstructured Homeopathic Case Sheet Documents Using Natural Language Processing Techniques

Abstract

Homeopathic physicians generally maintain the records of patients in the form of text documents, which contain valuable clinical information but lacks the standardized structure. Generation of structured data from unstructured homeopathic case sheet documents is a challenging process. The goal of the study is to identify and extract the key-value pairs from such unstructured documents to generate structured data that can be utilized for analysis and decision making in homeopathic practice. The paper outlines the methodology for collecting homeopathic case sheet documents, extracting relevant information and structuring it into a format suitable for further analysis. The methodology described in this study has been successfully implemented on a dataset comprising of 300 homeopathic case sheet documents. For each of these documents, our approach successfully extracted key-value pairs, representing various clinical parameters such as details of presenting complaints, family history, final diagnosis etc. The key-value pairs were structured and organized within excel spreadsheet, facilitating further analysis and interpretation of extracted clinical data. Evaluation shows that out of 300 documents three documents could not be properly converted to the structured form due to spacing constraints hence accuracy is 99%.