The Unit Graphs Framework: Foundational Concepts and Semantic Consequence

Proceedings of Recent Advances in Natural Language Processing, pages 389–395, Hissar, Bulgaria, 7-13 September 2013. The Unit Graphs Framework: Foundational Concepts and Semantic Consequence Maxime Lefranc¸ois and Fabien Gandon Wimmics, Inria, I3S, CNRS, UNSA 2004 rte des Lucioles, BP. 93, 06902 Sophia Antipolis, France {maxime.lefrancois,fabien.gandon}@inria.fr Abstract We are interested in a graph-based Knowl- edge Representation formalism that would allow for the representation, manipula- tion, query, and reasoning over depen- dency structures, and linguistic knowl- edge of the Explanatory and Combinato- rial Dictionary in the Meaning-Text The- ory framework. Neither the semantic web formalisms nor the conceptual graphs ap- pear to be suitable for this task, and this led to the introduction of the new Unit Graphs framework. This paper first in- troduces the foundational concepts of this framework: Unit Graphs are defined over a support that contains: i) a hierarchy of unit types which is strongly driven by their actantial structure, ii) a hierarchy of cir- cumstantial symbols, and iii) a set of unit identifiers. Then, this paper provides all of these objects with a model semantics that enables to define the notion of semantic consequence between Unit Graphs. 1 Introduction We are interested in the ability to reason over dependency structures and linguistic knowledge of the Explanatory and Combinatorial Dictionary (ECD), which is the lexicon at the core of the Meaning-Text Theory (MTT) (Mel’ˇcuk, 2006). Some formalisation works have been led on the ECD. For instance Kahane and Polgu`ere, 2001) proposed a formalization of Lexical Functions, and the Definiens project (Barque and Polgu`ere, 2008; Barque et al., 2010) aims at formalizing lex- icographic definitions with genus and specific dif- ferences for the TLFi1. Adding to these formaliza- tion works, the goal of the Unit Graphs formalism 1Tr´esor de la Langue Franc¸aise informatis´e, http:// atilf.atilf.fr is to propose a formalization from a knowledge engineering perspective, compatible with standard Knowledge Representation (KR) formalisms. The term formalization here means not only make non- ambiguous, but also make operational, i.e., such that it supports logical operations (e.g., knowl- edge manipulation, query, reasoning). We thus adopt a knowledge engineering approach applied to the domain of the MTT. At first sight, two existing KR formalisms seemed interesting for representing dependency structures: semantic web formalisms (RDF/S, OWL, SPARQL), and Conceptual Graphs (CGs) (Sowa, 1984; Chein and Mugnier, 2008). Both formalisms are based on directed labelled graph structures, and some research has been done to- wards using them to represent dependency struc- tures and knowledge of the lexicon (OWL in (Lefranc¸ois and Gandon, 2011; Boguslavsky, 2011), CGs at the conceptual level in (Bohnet and Wanner, 2010)). Yet Lefranc¸ois, 2013) showed that neither of these KR formalisms can repre- sent linguistic predicates. As the CG formalism is the closest to the semantic networks, the following choice has been made (Lefranc¸ois, 2013): Modify the CGs formalism basis, and define transforma- tions to the RDF syntax for sharing, and query- ing knowledge. As we are to represents linguistic units of different nature (e.g., semantic units, lexi- cal units, grammatical units, …

Content truncated to 500 words