The Event and Implied Situation Ontology (ESO): Application and Evaluation

The Event and Implied Situation Ontology (ESO): Application and Evaluation Roxane Segers⇤, Marco Rospocher ⌥, Piek Vossen ⇤, Egoitz Laparra †, German Rigau †, Anne-Lyse Minard⌥ ⇤Vrije Universiteit Amsterdam, The Netherlands {r.h.segers, p.t.j.m.vossen}@vu.nl ⌥Fondazione Bruno Kessler, Trento, Italy {rospocher, minard,}@fbk.eu † University of the Basque Country (UPV/EHU), Spain {egoitz.laparra, german.rigau}@ehu.eus Abstract This paper presents the Event and Implied Situation Ontology (ESO), a manually constructed resource which formalizes the pre and post situations of events and the roles of the entities affected by an event. The ontology is built on top of existing resources such as WordNet, SUMO and FrameNet. The ontology is injected to the Predicate Matrix, a resource that integrates predicate and role information from amongst others FrameNet, VerbNet, PropBank, NomBank and WordNet. We illustrate how these resources are used on large document collections to detect information that otherwise would have remained implicit. The ontology is evaluated on two aspects: recall and precision based on a manually annotated corpus and secondly, on the quality of the knowledge inferred by the situation assertions in the ontology. Evaluation results on the quality of the system show that 50% of the events typed and enriched with ESO assertions are correct. Keywords: Ontology, Semantic Role Labeling, Text Mining, Semantic Web 1. Introduction In this paper, we present the Event and Implied Situation Ontology (ESO) Version 2, that is matched with the Pred- icate Matrix (PM). Both resources rely on Semantic Role Labeling (SRL) descriptions and are used to detect and ab- stract over events, their participants and event implications in a large document collection about ten years of global au- tomotive industries, thus favoring the construction of large event-centric knowledge graphs (Rospocher et al., to ap- pear). ESO Version 2 (Segers et al., 2016) is a newly developed domain ontology to enhance the extraction and linking of dynamic and static events and their implications in text. Such a chain of changes and states and their implied situa- tions is presented in Figure 1. Here, the boxes represent var- ious event expressions about John’s employment while the ovals represent the implied situations of each event. Model- ing of event implications allows for extracting sequences of states and changes over time regardless of this information being directly expressed in text, or inferred by a reasoner. The model targets interpretations of situations rather than the semantics of predicates per se. Events are interpreted as situations using RDF, taking all event components into ac- count. Hence, the ontology and the linked resources need to be considered from the perspective of this interpretation model. Lexicons that define implications of events, for example VerbNet (Kipper et al., 2006), are rare and usually focus on the meaning of verbs in isolation. However, lexical struc- tures do not make explicit how the meaning of a verb needs to be combined with other event components, such as the participants and the temporal properties for the purpose of semantic parsing. We therefore follow an ontological ap- proach to interpret situations on the basis …

Content truncated to 500 words