In order to explain the cultural differences reported in the results of false-belief tasks, we attempted to verify the 'task bias hypothesis' suggested by certain studies (e.g. Tardif et al. (2004). Journal of Child Language, 31, 779-800; Rubio-Fernandez & Geurts (2013). Psychological Science, 24(1), 27-33. doi 10.1177/0956797612447819). At the same time, we aimed to observe the theory of mind (ToM) ability of infants and young children under the age of three in verbal communication. To this end, we propose a new protocol to test young children's ToM ability, with particular attention paid to the linguistic aspect of the task. This original disambiguation task using proper nouns (first names) was tested on a total of 32 children aged between 16 and 38 months, in France and Japan. The results revealed that after the age of 30 months children begin to correctly interpret nouns while simultaneously taking into account their partner's knowledge (50% of the French and 29% of the Japanese children were successful), whereas this remains difficult for younger children (no child under 30 months was successful). The analysis of error types has shown that 'memory bias' was dominant in younger children in particular and 'association bias' was rarely observed across all ages. Given that the results of French and Japanese children did not differ significantly, we assume that this new task design could minimise the influence of cultural difference caused by the characteristics of different languages.

: This article presents the state of the art in NLP research applied to Japanese and the relationship between this field and Japanese linguistics. Three aspects will be tackled : 1) the problems that are specific to NLP and that may trigger original reflection in the field of linguistics ; 2) the difficulties which linguists are not usually conscious of, but on which NLP casts new light ; 3) the data which are extensively dealt with in « traditional » linguistics and which are specific to Japanese — and understanding the latter requires the study of the Japanese linguistics' literature.

No abstract

In Japanese, as for many other languages, quantification of countable nouns is realized with numeral phrases, consisting of the numeral and the c1assifier. Each noun has a restricted set of possible c1assifiers from an inventory of approximately 100. The association between noun and classifier is threefold: It can be of either the lexical, semantic, or morphological type. This article proposes a formalization of the numeral phrase-noun association in order to analyse them with a syntactic-semantic parser. The present work is part of a larger research project on the formal syntax and semantics of Japanese numeral constructions.