Vietnamese Electronic Medical Record Management with Text Preprocessing for Spelling Errors

Abstract

With an important development of information technology, electronic medical records (EMRs) play a crucial role in medicine and its related fields such as pharmacology, bioinformatics, and healthcare research. An EMR management system is thus significant so that EMRs can be manipulated conveniently by their key users, who are doctors, nurses, and patients, and further exploited by other researchers. Nonetheless, there is no consideration on the EMRs when they are committed to the system. Their clinical texts are especially left behind upon their entry. Such a circumstance is understandable because they are created in a high-pressure working environment and normally late examined as needed in other tasks. In order to fill this gap, we propose an EMR management system with text preprocessing for spelling error correction on clinical texts. Our system is dedicated to Vietnamese EMRs with a practical demonstration confirmed by doctors and medical students. In addition, the spelling error correction method is defined to cover a wide diversity of spelling errors in a novel hybrid manner, by effectively combining a rule-based approach, dictionaries, n-grams, and a pre-trained monolingual sequence-to-sequence model. Indeed, the experimental results show that the method has better performance on real EMRs as compared to one of the most recent existing ones. As a result, it is promising that our work can enhance and make Vietnamese EMRs more utilizable.