PageRanking WordNet Synsets: An Application to Opinion Mining

Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics, pages 424–431, Prague, Czech Republic, June 2007. c⃝2007 Association for Computational Linguistics PageRanking WordNet Synsets: An Application to Opinion Mining∗ Andrea Esuli and Fabrizio Sebastiani Istituto di Scienza e Tecnologie dell’Informazione Consiglio Nazionale delle Ricerche Via Giuseppe Moruzzi, 1 – 56124 Pisa, Italy {andrea.esuli,fabrizio.sebastiani}@isti.cnr.it Abstract This paper presents an application of PageR- ank, a random-walk model originally de- vised for ranking Web search results, to ranking WordNet synsets in terms of how strongly they possess a given semantic prop- erty. The semantic properties we use for ex- emplifying the approach are positivity and negativity, two properties of central impor- tance in sentiment analysis. The idea derives from the observation that WordNet may be seen as a graph in which synsets are con- nected through the binary relation “a term belonging to synset sk occurs in the gloss of synset si”, and on the hypothesis that this relation may be viewed as a transmit- ter of such semantic properties. The data for this relation can be obtained from eX- tended WordNet, a publicly available sense- disambiguated version of WordNet. We ar- gue that this relation is structurally akin to the relation between hyperlinked Web pages, and thus lends itself to PageRank analysis. We report experimental results supporting our intuitions. 1 Introduction Recent years have witnessed an explosion of work on opinion mining (aka sentiment analysis), the dis- ∗This work was partially supported by Project ONTOTEXT “From Text to Knowledge for the Semantic Web”, funded by the Provincia Autonoma di Trento under the 2004–2006 “Fondo Unico per la Ricerca” funding scheme. cipline that deals with the quantitative and qualita- tive analysis of text for the purpose of determining its opinion-related properties (ORPs). An important part of this research has been the work on the auto- matic determination of the ORPs of terms, as e.g., in determining whether an adjective tends to give a positive, a negative, or a neutral nature to the noun phrase it appears in. While many works (Esuli and Sebastiani, 2005; Hatzivassiloglou and McKeown, 1997; Kamps et al., 2004; Takamura et al., 2005; Turney and Littman, 2003) view the properties of positivity and negativity as categorical (i.e., a term is either positive or it is not), others (Andreevskaia and Bergler, 2006b; Grefenstette et al., 2006; Kim and Hovy, 2004; Subasic and Huettner, 2001) view them as graded (i.e., a term may be positive to a certain degree), with the underlying interpretation varying from fuzzy to probabilistic. Some authors go a step further and attach these properties not to terms but to term senses (typ- ically: WordNet synsets), on the assumption that different senses of the same term may have dif- ferent opinion-related properties (Andreevskaia and Bergler, 2006a; Esuli and Sebastiani, 2006b; Ide, 2006; Wiebe and Mihalcea, 2006). In this paper we contribute to this latter literature with a novel method for ranking the entire set of WordNet synsets, irrespectively of POS, according to their ORPs. Two rankings …

Content truncated to 500 words