Computational Modeling of Verb Acquisition, from a Monolingual to a Bilingual Study

Computational modeling of verb acquisition, from a monolingual to a bilingual study ⋆ Laurent Prévot a, Chun-Han Chang b, and Yann Desalle c aLPL, CNRS - Université de Provence, France laurent.prevot@lpl-aix.fr bGraduate Institute of Linguistics, National Taiwan University, Taiwan chunhanchang@ntu.edu.tw cOctogone, CNRS - Université de Toulouse, France desalle@univ-tlse2.fr Abstract. Going cross-linguistic is a an important but challenging track for validating a com- putational model of lexical organization. Our starting point is a computational model that has been established and validated on French language and we attempted to apply it on Man- darin language. The main ingredients of this model are computational lexical resources and a psycho-linguistic protocol involving extra-linguistic material (video-clips). At this stage, all the psycho-linguistic experiments have been ran, most of the resources have been built but some comparative analyses are not fully completed. Still the project is advanced enough to report on the issues we had to address while performing this cross-linguistic move concerning the resources, the analysis of the data and the data alignment across languages. Keywords: Computational model, Acquisition, cross-linguistic study, French, Mandarin 1 Introduction The study reported in this paper is part of a broader project (M3- Model and Measurement of Meaning) which investigates lexical organization of French and Mandarin.1 The project combines psycho-linguistic and computational approaches for investigating the verb lexical organization of both languages. We therefore deal with two kinds of data: On the one hand the productions obtained through psycho-linguistic experiments performed with French and Mandarin Chinese speakers; On the other hand we systematically exploit existing electronic resources. A schematic organization of the project in proposed in Figure 1. Two parallel cross-linguistic studies were planned. First at the psycho-linguistic level, the same protocol, Approx (Duvignau and Gaume, 2003; Duvignau et al., 2005; Duvignau et al., 2007)), for analyzing the productions at different ages (Pi.fr/Pi.tw = French/Taiwanese Population of Age Group i,...) would be applied to both languages for studying productions variability across languages. Then at a computational level, existing lexical resources, in our case paradigmatic graphs2 (PG.fr/PG.tw = French paradigmatic graph, Mandarin paradigmatic graph), should be aligned thanks to translation resources in order to compare the structural properties and the organization of the graphs. For example, we would like to check whether some lexical clusters in a language are preserved when projected in the other language and if it is the case which ones. ⋆ The research presented in this paper is supported by the ANR-NSC (France-Taiwan) bilateral project M3 (Modeling and Measurement of Meaning). Copyright 2010 by Laurent Prévot, Chun-Han Chang, and Yann Desalle 1 More information about M3 project on these websites: http://erss.irit.fr/flexsem (France) http://140.112.147.149:81/m3 /(Taiwan). 2 Paradigmatic graphs a here a sort of synonymy graph in which the synonymy relation is kept relatively loose. PACLIC 24 Proceedings 841 Figure 1: M3 Square A validation across the methodologies is proposed as the following. Given a set of answers of a population, we can look at their properties in the paradigmatic graph. Then as explained in this volume (Desalle et …

Content truncated to 500 words