Analysis of patient reviews using machine learning and linguistic methods

With the advancement of digitalization, traditional methods of surveying consumers to assess their satisfaction with service quality are being replaced by approaches based on the automatic processing of text data from social media. This study aims to determine the degree of patient satisfaction with the quality of medical services by developing and testing an algorithm for classifying Russian-language text reviews collected from social media platforms. The focus is on analyzing the sentiment (positive/negative) of patient reviews about medical institutions and doctors, as well as identifying the review's subject-either the quality of medical services provided or the organization of patient care by the institution. A method was developed for classifying text reviews about the work of medical institutions posted by patients on two Russian doctor review platforms. Approximately 60,000 reviews were analyzed. Machine learning techniques, including various artificial neural network architectures, were tested. The classification algorithm demonstrated high efficiency, with the best performance achieved using a recurrent neural network architecture (accuracy = 0.9271). Incorporating named entity recognition into text analysis further enhanced the classification efficiency across all neural network-based classifiers. To improve classification quality, the study highlights the need for semantic segmentation of reviews by their subject and sentiment, followed by the separate analysis of these fragments.