Adaptation of Semantic Rule-Based Sentiment Analysis Approach for Russian Language

The paper describes application of the semantic rule-based sentiment analysis approach, which was earlier developed and tested on English texts, to the Russian language. In order to take into account specificity of Russian it was adapted, particularly representation of the rules as patterns over a list of words was replaced with algorithms over the syntax tree of a sentence. The experiments on a quarter of a corpus of sentences extracted from hotel reviews allowed to perform the error analysis and refinement of the approach. The final results on the whole corpus allowed to achieve the results close to the state-of-the-art methods based on neural networks. The advantages of the approach, including simple interpretability of its results and absence of the need of learning, make it perspective for further research in sentiment analysis.