Details

The target of the research are user reviews on medical institutions’ services aggregated on the “Yandex-Maps” website. The scope of the research are methods for classifying natural language texts by sentiment and methods for modeling major topics of these texts. The objective of the study is to collect data to train a chosen sentiment analysis algorithm and model the main topics of user reviews. The tasks of the research: 1) To collect and analyze the scientific literature and technical documentation on the master thesis topic; 2) To develop and implement a program for extracting user reviews aggregated on the Yandex-Map website; 3) To build a corpus of user reviews on medical institutions’ services, to analyze it and preprocess it; 4) To train the ML models on the compiled corpus and identify the most effective one for the task of sentiment analysis; 5) To train the topic modelling algorithm on the compiled corpus; 6) To analyze the results of topic modelling and obtained key topics and the results of sentiment analysis. The research methodologies include descriptive method, comparative method, statistical method, experiment, and machine learning method. As a result of the study a program for automatic collection of user reviews from the “Yandex-Maps” website was created, and a corpus of reviews to medical facilities in Russian was collected. Machine learning algorithms for sentiment analysis and algorithms for topic modelling were trained and compared. Analysis of user reviews provides an opportunity to identify the main disadvantages and advantages concerning a product or service in any domain.