How do children acquire early grammar and build multiword utterances? A corpus study of French children aged 2 to 4.

In the last 50 years, researchers have debated over the lexical or grammatical nature of children's early multiword utterances. Due to methodological limitations, the issue remains controversial. This corpus study explores the effect of grammatical, lexical, and pragmatic categories on mean length of utterances (MLU). A total of 312 speech samples from high-low socioeconomic status (SES) French-speaking children aged 2-4 years were annotated with a part-of-speech-tagger. Multiple regression analyses show that grammatical categories, particularly the most frequent subcategories, were the best predictors of MLU both across age and SES groups. These findings support the view that early language learning is guided by grammatical rather than by lexical words. This corpus research design can be used for future cross-linguistic and cross-pathology studies.

Automatic disambiguation of morphosyntax in spoken language corpora.

The use of computer tools has led to major advances in the study of spoken language corpora. One area that has shown particular progress is the study of child language development. Although it is now easy to lexically tag every word in a spoken language corpus, one still has to choose between numerous ambiguous forms, especially with languages such as French or English, where more than 70% of words are ambiguous. Computational linguistics can now provide a fully automatic disambiguation of lexical tags. The tool presented here (POST) can tag and disambiguate a large text in a few seconds. This tool complements systems dealing with language transcription and suggests further theoretical developments in the assessment of the status of morphosyntax in spoken language corpora. The program currently works for French and English, but it can be easily adapted for use with other languages. The analysis and computation of a corpus produced by normal French children 2-4 years of age, as well as of a sample corpus produced by French SLI children, are given as examples.

How children build their morphosyntax: the case of French.

Early morphosyntax is very rich and uniform in young French-speaking children. The present study aims to give a thorough analysis of the morphosyntax produced at the outset of multi-word speech, with a classification of free language produced at 2;0 by 27 French-speaking children. The corpus was fully tagged by an automatic part-of-speech tagger. A classification performed with words taken in isolation shows a clear difference between the categories used in single-word utterances and those used in multi-word utterances. A classification performed with word sequences reveals surprisingly adult-like sequences of syntactic categories and words; the non-adult combinations are few in a French child's language. The very successful use of the tagger demonstrates the morphosyntactic coherence of the child's speech. When compared with adult language, the quantitative results, and more precisely the data concerning regularity and error types, contribute to the documentation of all the specificities of the emerging morphosyntax in normally developing French children.