A Distributional Model of Verb-Specific Semantic Roles Inferences

In a standard view, commonly adopted in psycholinguistics and computational linguistics, thematic roles are approached as primitive entities able to represent the roles played by the arguments of a pred- icate. In theoretical linguistics, however, the inability to reach a con- sensus on a primitive set of semantic roles led to the proposal of new approaches in which thematic roles are better described as a bundle of more primitive entities (e.g., Dowty, 1991; Van Valin, 1999) or as structural configurations (e.g., Jackendoff, 1987). In a complemen- tary way, psycholinguistic evidence supports the idea that thematic roles and nominal concepts are represented in similar ways (McRae et al., 1997b; Ferretti et al., 2001), thus suggesting that the former can be accounted for as predicate-specific bundles of inferences activated by the semantics of the verb (e.g., the patient of kill is typically alive before the event and dead afterward). Such inferences can take the form of either presuppositions or entailment relations activated when a filler saturates a specific argument position for a given predicate. Our aim in this chapter is twofold. First, we report behavioral data collected to obtain a more fine-grained characterization of the the- matic role properties activated by a subset of English verbs. To this end, we employed the modified version of the McRae et al. (1997b) elicitation paradigm proposed by Lebani et al. (2015) to describe which semantic properties of the participants are more relevant in each phase of the action described by the predicate. Next, we test the possibility to model such verb-specific inference patterns by exploit- ing corpus-based distributional data, thus proposing a novel approach to represent the same level of semantic knowledge that is currently described by means of a finite set of thematic roles