Aspect Extraction and Sentiment Analysis in User Reviews in Russian about Bank Service Quality

The paper considers the problem of automated extraction of aspect terms from user reviews in Russian about bank service quality. The purpose of this research is to develop a method for the sentiment analysis within the bank service quality domain. The sentiment lexicon is automatically extracted together with aspect words using a rule-based algorithm and Pointwise Mutual Information. The sentiment lexicon containing 460 positive, 538 negative words and 69 aspect word is extracted. This lexicon is compared with the one manually built on the same reviews and the efficiency is estimated with Precision, Recall, and F1. The suggested method of automated extraction of sentiment words demonstrated sufficient Precision and Recall values, especially for high-frequency words.