Rationale, Concepts, and Current Outcome of the Unit Graphs Framework

Proceedings of Recent Advances in Natural Language Processing, pages 382–388, Hissar, Bulgaria, 7-13 September 2013. Rationale, Concepts, and Current Outcome of the Unit Graphs Framework 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 The Unit Graphs (UGs) framework is a graph-based knowledge representation (KR) formalism that is designed to al- low for the representation, manipula- tion, query, and reasoning over linguis- tic knowledge of the Explanatory Com- binatorial Dictionary of the Meaning-Text Theory (MTT). This paper introduces the UGs framework, and overviews current published outcomes. It first introduces rationale of this new formalism: nei- ther semantic web formalisms nor Con- ceptual Graphs can represent linguistic predicates. It then overviews the foun- dational concepts of this framework: the UGs are defined over a UG-support that contains: i) a hierarchy of unit types which is strongly driven by the actantial struc- ture of unit types, ii) a hierarchy of cir- cumstantial symbols, and iii) a set of unit identifiers. On these foundational con- cepts and on the definition of UGs, this paper finally overviews current outcomes of the UGs framework: the definition of a deep-semantic representation level for the MTT, representation of lexicographic def- initions of lexical units in the form of se- mantic graphs, and two formal semantics: one based on UGs closure and homomor- phism, and one based on model semantics. 1 Introduction The Meaning-Text Theory (MTT) is a theoreti- cal dependency linguistics framework for the con- struction of models of natural language. As such, its goal is to write systems of explicit rules that ex- press the correspondence between meanings and texts (or sounds) in various languages (Kahane, 2003). From semantic representations to surface phonologic representations, seven different levels of linguistic representation are supposed for each set of synonymous utterances. Thus, two times six modules containing transformation rules are used to transcribe representations of a level into representations of an adjacent level. The main constituent of the MTT is the dictionary model where lexical units are described, which is called the Explanatory Combinatorial Dictionary (ECD) (Mel’ˇcuk, 2006). As for any community of interest, linguists and lexicographers of the MTT framework produce knowledge. Knowledge Representation (KR) is an area of artificial intelligence that deals with recur- rent needs that emerge with such knowledge pro- duction. The aim of this paper is to introduce the Unit Graphs KR formalism that is designed to allow for the representation, manipulation, query, and rea- soning over dependency structures, rules and lexi- cographic knowledge of the ECD. The rest of this paper is organized as follows. We will first introduce rationale of this new KR formalism (§2), then the fundamental concepts of the UGs framework (§3), implications for the MTT, lexicographic definitions and application to a specific MTT lexicographic edition project (§4), and finally two approaches to assign UGs with log- ical semantics, so as to enable reasoning in the UGs framework (§5). 2 Rationale: Representation of Valency-based Predicates Most past …

Content truncated to 500 words