RusEmoLex: Russian Emotion Lexicon

Emotion detection in natural language is an important area of computational linguistics. It involves the description, categorisation and systematization of emotion lexicon that can be further employed to enhance the performance of machine learning algorithms. The aim of the study is twofold: firstly, to provide a comprehensive description of the Russian-language resources pertaining to emotion vocabulary, and secondly, to create an emotion lexicon based on these resources. The study describes the following types of Russian-language sources: lexicographic resources (dictionaries and thesauri); data from the semantically tagged part of the Russian National Corpus; datasets consisting of texts labelled by emotion type; and word-emotion association lexicons. The methodology of creating RusEmoLex encompasses the following stages: the creation of initial lists of emotion words derived from available sources; the compilation of the lists; and the selection of lexis in RusEmoLex. A lexical unit has to meet the following criteria to be included in RusEmoLex: repeated occurrence in the sources (a lexical unit needs to be present in two or more sources); the quality of the sources (a lexical unit needs to be included either in a dictionary or a corpus resource and it should not be a translation); the class label (the lexical units have an emotional label established on the basis of the majority of sources). RusEmoLex is an open access resource which includes 1024 lexical units and can be used for linguistic studies and in the field of natural language processing.