PreMOn: a Lemon Extension for Exposing Predicate Models as Linked Data

PreMOn: a Lemon Extension for Exposing Predicate Models as Linked Data Francesco Corcoglioniti, Marco Rospocher, Alessio Palmero Aprosio, Sara Tonelli Fondazione Bruno Kessler – IRST Via Sommarive 18, Trento, I-38123, Italy {corcoglio,rospocher,aprosio,satonelli}@fbk.eu Abstract We introduce PreMOn (predicate model for ontologies), a linguistic resource for exposing predicate models (PropBank, NomBank, VerbNet, and FrameNet) and mappings between them (e.g, SemLink) as Linked Open Data. It consists of two components: (i) the PreMOn Ontology, an extension of the lemon model by the W3C Ontology-Lexica Community Group, that enables to homogeneously represent data from the various predicate models; and, (ii) the PreMOn Dataset, a collection of RDF datasets integrating various versions of the aforementioned predicate models and mapping resources. PreMOn is freely available and accessible online in different ways, including through a dedicated SPARQL endpoint. Keywords: Predicate Models; Predicate Model Mappings; Lemon; Semantic Web; Linguistic Linked Open Data. 1. Introduction Predicate models such as PropBank (herafter, PB) (Palmer et al., 2005), NomBank (NB) (Meyers et al., 2004), Verb- Net (VN) (Schuler, 2005), and FrameNet (FN) (Baker et al., 1998) provide rich descriptions of predicate semantic classes — i.e., rolesets in NB and PB, verb classes in VN, and frames in FN (e.g., “Commerce Sell”) — and their se- mantic roles (e.g., “Seller” and “Buyer”), abstracting from a number of linguistic phenomena related to their realiza- tion in text. Thanks to the mappings of different predicate models, such as SemLink (Palmer, 2009) and the Predi- cate Matrix (Lacalle et al., 2014), and to their integration in Semantic Role Labeling (SRL) tools, they have become central to a number of tasks such as information extrac- tion, question answering and natural language generation. In particular, due to their laying at the syntactic-semantics interface, predicate models are increasingly used within the Semantic Web (SW) community, for knowledge extraction in tools such as NewsReader (Rospocher et al., 2016) and PIKES (Corcoglioniti et al., 2016), or as the starting point for deriving general-domain ontologies grounded in natural language, such as FrameBase (Rouces et al., 2015) and the ESO ontology (Segers et al., 2015) derived from FN. Compared to the current situation where each predicate model has its own terminology, structure and proprietary XML format, the availability of a single RDF/OWL on- tological model covering the main predicate models with their common aspects, specificities, and mappings would be beneficial to all the applications mentioned above, within and outside the SW area. Indeed, the use of SW tech- nologies and the adoption of the Linked Open Data (LOD) paradigm have already been recognized as particularly ben- eficial to linguistic resources (see, e.g., Chiarcos et al. (2013)), leading to the creation of the Linguistic Linked Open Data cloud curated by the Open Linguistic subgroup of the Open Knowledge Foundation.1 In that context, the lemon (lexical model for ontologies) model by McCrae et al. (2012), recently revised by the W3C Ontology Lexicon Community Group (2015), already sets the basis for model- ing the lexical entries referenced in predicate models (e.g., 1http://linguistics.okfn.org/ the verb ‘sell’), …

Content truncated to 500 words