ÆTHEL: Automatically Extracted Typelogical Derivations for Dutch

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 5257–5266 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 5257 ÆTHEL: Automatically Extracted Typelogical Derivations for Dutch Konstantinos Kogkalidis , Michael Moortgat , Richard Moot Utrecht Institute of Linguistics OTS, Utrecht University, LIRMM, Universit´e de Montpellier, CNRS {k.kogkalidis, m.j.moortgat}@uu.nl, richard.moot@lirmm.fr Abstract We present ÆTHEL, a semantic compositionality dataset for written Dutch. ÆTHEL consists of two parts. First, it contains a lexicon of supertags for about 900 000 words in context. The supertags correspond to types of the simply typed linear lambda-calculus, enhanced with dependency decorations that capture grammatical roles supplementary to function-argument structures. On the basis of these types, ÆTHEL further provides 72 192 validated derivations, presented in four equivalent formats: natural-deduction and sequent-style proofs, linear logic proofnets and the associated programs (lambda terms) for meaning composition. ÆTHEL’s types and derivations are obtained by means of an extraction algorithm applied to the syntactic analyses of Lassy Small, the gold standard corpus of written Dutch. We discuss the extraction algorithm and show how ‘virtual elements’ in the original Lassy annotation of unbounded dependencies and coordination phenomena give rise to higher-order types. We suggest some example usecases highlighting the benefits of a type-driven approach at the syntax semantics interface. The following resources are open-sourced with ÆTHEL: the lexical mappings between words and types, a subset of the dataset consisting of 7 924 semantic parses, and the Python code that implements the extraction algorithm. Keywords: Categorial Grammar, Grammar Extraction, Dutch Treebank 1. Introduction Typelogical categorial grammars offer an attractive view on the syntax-semantics interface. The key idea is that the derivation establishing the well-formedness of a phrase systematically corresponds to a procedure that computes the meaning of that phrase, given the meanings of its con- stituent words. This connection between logical derivations and programs, known as the Curry-Howard correspondence (Sørensen and Urzyczyn, 2006), plays a central role in current typelog- ical grammars. In extended versions of the Lambek cal- culus (Lambek, 1958) such as (Morrill, 1994; Moortgat, 1996; Morrill et al., 2011; Kubota and Levine, to appear), the Curry-Howard correspondence applies to a language of semantic types, where meaning composition is addressed. The semantic type language is obtained by a structure- preserving translation from the syntactic type system that handles surface word order, prosodic structure and the like. Alternatively, the type system of Abstract Categorial Gram- mar (De Groote, 2001) and Lambda Grammars (Muskens, 2001) is designed to capture a ‘tectogrammatical’ (Curry, 1961) view of language structure that abstracts away from the surface realization. From the abstract syntax now both the meaning representations and the surface forms are ob- tained by compositional translations, where for the latter one relies on the standard encoding of strings, trees, etc in lambda calculus (Barendregt et al., 2013). A common feature of these approaches is the use of resource-sensitive type logics, with Linear Logic (Girard, 1987) being the key representative. Curry’s original for- mulation of the logic-computation correspondence was …

Content truncated to 500 words