Automatically Extracting Qualia Relations for the Rich Event Ontology

Proceedings of the 27th International Conference on Computational Linguistics, pages 2644–2652 Santa Fe, New Mexico, USA, August 20-26, 2018. 2644 Automatically Extracting Qualia Relations for the Rich Event Ontology Ghazaleh Kazeminejad1, Claire Bonial2, Susan Windisch Brown1 and Martha Palmer1 1University of Colorado Boulder, 2U.S. Army Research Lab {ghazaleh.kazeminejad, susan.brown, martha.palmer}@colorado.edu claire.n.bonial.civ@mail.mil Abstract Commonsense, real-world knowledge about the events that entities or “things in the world” are typically involved in, as well as part-whole relationships, is valuable for allowing computational systems to draw everyday inferences about the world. Here, we focus on automatically extracting information about (1) the events that typically bring about certain entities (origins), (2) the events that are the typical functions of entities, and (3) part-whole relationships in entities. These correspond to the agentive, telic and constitutive qualia central to the Generative Lexicon. We describe our motivations and methods for extracting these qualia relations from the Suggested Upper Merged Ontology (SUMO) and show that human annotators overwhelmingly find the information extracted to be reasonable. Because ontologies provide a way of structuring this information and making it accessible to agents and computational systems generally, efforts are underway to incorporate the extracted information to an ontology hub of Natural Language Processing semantic role labeling resources, the Rich Event Ontology. 1 Introduction Most adults could tell you that an umbrella is used to keep the rain off of you, that bread is baked, or that dandelions are made up of leaves, stems, and flowers. Commonsense, real-world knowledge about the events that entities or “things in the world” are typically involved in, as well as part-whole relationships, is valuable for allowing computational systems to draw everyday inferences about the world. Here, we describe our approach for automatically extracting this information from the Suggested Upper Merged Ontology (SUMO) (Niles and Pease, 2001a; Pease, 2011). We assume the theoretical framework of the Generative Lexicon (GL) (Pustejovsky, 1991), as we find GL qualia relations useful for considering the typical involvement of a particular entity in events. Thus, we focus on relationships between entities and events captured in a set of qualia associated with an entity. We find that human annotators asked to evaluate the quality and accuracy of the information extracted overwhelmingly find the origins, functions and part-whole relationships proposed to be reasonable. One of our motivations is to add the extracted qualia relations to an ontology hub of Natural Language Processing (NLP) semantic role labeling (SRL) resources, the Rich Event Ontology (REO). We plan to exploit the qualia relations within REO to make generalizations across participant types for a given event type, thereby facilitating the addition of selectional restrictions, and to enable some shallow inferencing about how entities generally come into existence, how they are used, and what they are made up of. By structuring this information within REO we aim, for example, to allow an agent or computational system to infer that a room (whether detected visually or described in text) with a stove, pans, and a refrigerator is likely a kitchen where …

Content truncated to 500 words