Creating a General Russian Sentiment Lexicon

Creating a General Russian Sentiment Lexicon Natalia Loukachevitch, Anatoly Levchik Lomonosov Moscow State University, RJ games Moscow, Russia E-mail: louk_nat@mail.ru, endight@gmail.com Abstract The paper describes the new Russian sentiment lexicon - RuSentiLex. The lexicon was gathered from several sources: opinionated words from domain-oriented Russian sentiment vocabularies, slang and curse words extracted from Twitter, objective words with positive or negative connotations from a news collection. The words in the lexicon having different sentiment orientations in specific senses are linked to appropriate concepts of the thesaurus of Russian language RuThes. All lexicon entries are classified according to four sentiment categories and three sources of sentiment (opinion, emotion, or fact). The lexicon can serve as the first version for the construction of domain-specific sentiment lexicons or be used for feature generation in machine-learning approaches. In this role, the RuSentiLex lexicon was utilized by the participants of the SentiRuEval-2016 Twitter reputation monitoring shared task and allowed them to achieve high results. Keywords: sentiment analysis, sentiment lexicon, connotation, lexical ambiguity, thesaurus 1. Introduction Automatic sentiment analysis is useful in many practical applications, such as analysis of users' reviews, posts in social networks, newspaper articles, etc. Sentiment lexicons are important components of sentiment analysis systems. They can be applied in lexicon-based approaches (Taboada et al., 2011) or be sources of features in the machine-learning framework (Mohammad et al., 2013; Severyn & Moschitti, 2015). For English, there are many sentiment lexicons, which were manually created by experts (Wilson et al., 2005) or by crowdsourcing (Mohammad & Turney, 2013). For other languages, automatic approaches for generating sentiment lexicons have been proposed (Chetviorkin & Loukachevitch, 2012; Perez-Rosas et al., 2012; San Vicente et al., 2014; Yang et al., 2013). It is well-known that sentiment vocabularies can depend on a specific domain, therefore, a lot of works are devoted to extraction of domain-specific sentiment lexicons (Blitzer et al., 2007; Choi & Cardie, 2009; Lau et al., 2011). But some works (Mansour et al., 2013) show that the combination of multi-domain training data in supervised sentiment analysis improves the performance of a classifier in each domain under study. It proves the existence of a relatively stable set of general sentiment words and expressions with relatively stable sentiment orientations. Besides, as was shown in (Mohammad et al., 2013), features based on publicly available sentiment vocabularies are useful for improvement of supervised sentiment analysis systems. Thus, for any natural language it is useful to have a publicly available, manually crafted general sentiment lexicon, which can be later adapted to specific domains. In this paper we present a new manually created general Russian Sentiment Lexicon − RuSentiLex. The lexicon contains more than ten thousand Russian sentiment-related words and expressions. Ambigous words that have different sentiment polarity in different senses are provided with links to appropriate concepts of the Thesaurus of Russian language RuThes (Loukachevitch & Dobrov, 2014), which can help disambiguate sentiment ambiguity in specific domains or contexts. The rest of this paper is organized as follows. In the second section we consider related …

Content truncated to 500 words