Acquiring Opposition Relations among Italian Verb Senses using Crowdsourcing

Acquiring Opposition Relations among Italian Verb Senses using Crowdsourcing Anna Feltracco1,2, Simone Magnolini1,3, Elisabetta Jezek2, Bernardo Magnini1 1Fondazione Bruno Kessler, Via Sommarive 18, 38100 Povo-Trento, Italy 2University of Pavia, Strada Nuova 65, 27100 Pavia, Italy 3University of Brescia, Piazza del Mercato 15, 25121 Brescia, Italy feltracco@fbk.eu, magnolini@fbk.eu, jezek@unipv.it, magnini@fbk.eu Abstract We describe an experiment for the acquisition of opposition relations among Italian verb senses, based on a crowdsourcing methodology. The goal of the experiment is to discuss whether the types of opposition we distinguish (i.e. complementarity, antonymy, converseness and reversiveness) are actually perceived by the crowd. In particular, we collect data for Italian by using the crowdsourcing platform CrowdFlower. We ask annotators to judge the type of opposition existing among pairs of sentences -previously judged as opposite- that differ only for a verb: the verb in the first sentence is opposite of the verb in second sentence. Data corroborate the hypothesis that some opposition relations exclude each other, while others interact, being recognized as compatible by the contributors. Keywords: opposition relations, verb sense, crowdsourcing, T-PAS resource 1. Introduction Traditionally, the linguistic study of semantic relations among verbs or verb senses has focused on the manner re- lation (also known as troponymy, e.g. move and walk), the cause relation (kill and die), and, more generally the re- lation of lexical entailment (snore and sleep) - which, ac- cording to the classification in Fellbaum (1998), subsumes several relation types (such as cause). In the computational field, several initiatives have proposed schemes for the an- notation of both the internal structure of the events encoded in verbs (see, for instance, Aguilar et al. (2014), Fokkens et al. (2013)) and the relations among events, including tem- poral relations as proposed in the TimeML scheme (Puste- jovsky et al., 2003). Less works have systematically ad- dressed the relation of opposition between verbs or verb senses (increase and decrease) and its annotation in exist- ing lexicons or sense repertoires. In this paper we describe an experiment we run to acquire opposition relations among Italian verb senses, based on a crowdsourcing methodology. In the experiment, we assume a four-output classification of opposition types, namely complementarity, antonymy, converseness and reversive- ness. We rely on an existing repository of verb frames for Ital- ian (see Section 5.1), and are interested to verify how the different opposition types are perceived by the crowd and whether it is possible for a pair of verb senses to have char- acteristics that belong to more than one type. The potential output of the work is a gold-standard of verb frame pairs annotated with opposition relations. The paper is organized as follows. Section 2 introduces our notion of opposition and the classification we use in our work and Section 3 presents examples of related works using the crowdsourcing methodology. In Section 4 the methodology we adopt is described, followed by the ex- perimental settings in Section 5. Finally, in Section 6 the results are discussed while Section 7 provides some con- clusions and directions …

Content truncated to 500 words