MultiGenera: working steps

At the Euralex2018 congress ( we had the opportunity to present the projects MultiGenera and MultiComb. This time we focus especially on explaining the linguistic and computational methodology of our projects for the automatic generation of nominal arguments, in particular we concentrate on explaining the MultiGenera project. The aim of this research is to develop a prototype for a generator of argument structure or valency realisations in terms of syntagmatic and paradigmatic combinations of Spanish, German and French nouns.

The multilingual tool prototype MultiGenera is designed according to the principles of Valency Grammar and Lexicography (Engel 1995), and following related work on WordNets, i.e. concept-based lexical resources represented as ontologies. This will allow an analysis of different lexical-semantic domains in Spanish, German, and French from an onomasiological-conceptual point of view. The two main applications of the tool prototype are: (1) the generation of noun phrases as argument structure realizations that follow patterns related to semantic features, for the creation of corpus and web query strings, and (2) the knowledge-based generation of simple and complex noun phrases that are acceptable in a coherent sentence context.

The design and development of MultiGenera and MultiComb follows five core working steps, the description of argument structure realisation patterns, expansion of lexical prototypes, the generation of argument structure realisations, argument combinations within noun phrases, and context generation. For more information see here.