Community Research and Development Information Service - CORDIS


ACQDIV Report Summary

Project ID: 615988
Funded under: FP7-IDEAS-ERC
Country: Switzerland

Periodic Report Summary 2 - ACQDIV (Acquisition processes in maximally diverse languages: Min(d)ing the ambient language)

The languages of the world are vastly different from each other and vary enormously in complexity. Notwithstanding, a child can learn any of them without difficulties. ACQDIV's main concern is why and how this is possible. Our leading hypothesis is that there is a small set of cognitive mechanisms that all humans are endowed with; these change during development and are not necessarily tailored to language acquisition. Some candidates for such mechanisms are the recognition and reproduction of abstract patterns, learning based on exemplars or rules abstracting over them, and learning in social interactions. In order to find out how universal these mechanisms really are, we are using longitudinal corpora from 10 languages that were selected for being maximally diverse in their grammatical structures: Chintang, Cree, Dënë Sųłıné, Indonesian, Inuktitut, Japanese, Russian, Sesotho, Turkish, and Yucatec. For nine of these we use pre-existing corpora (partly provided by collaborators); for one (Dënë Sųłıné, an endangered First Nations language of Canada) we compile a corpus ourselves.
Dënë Sųłıné belongs to one of the most complex language families in the world. The two communities we are working with for this subproject, La Loche and the Clearwater River Dene Nation, are both relatively remote. In the first 30 months of the project we established working relations with local institutions such as the Clearwater River Dene School. With their approval and help we found a sufficient number of collaborating families and children, as well as local field assistants for recording, transcribing and translating the language of children and surrounding adults in their natural environment. By now we have video-recorded 437 sessions for 12 children, amounting to more than 460 hours of data. Transcriptions, translations and analyses have started. In addition, ACQDIV is collaborating with the Dene Teachers Program (DTEP) and the Dene Language and Culture Committee to provide language support and help to compile a dictionary for use in schools.
The main technical achievement of the first 18 months was the design and development of a database containing all corpora in a unified format. This allows us to search through data from all our 10 languages simultaneously and automatically. The ACQDIV Corpus is the first data set of this kind in language acquisition research.
The initial research focus of the project was on input patterns, i.e. regular patterns that occur in child-surrounding speech and can be recognized and used by children for learning a language. One such pattern are frequent frames, which may help children to learn parts-of-speech. For instance, the combination of the two words "a" and "doggy" (one example for a frame) most often occurs with different adjectives in the middle (e.g. "nice") and therefore might help children to recognize adjectives as a distinct class of words and learn about their distributions. We found that frequent frames exist in child-surrounding speech in all languages in our sample for which we had sufficient data. However, in languages with high grammatical complexity this pattern only exists below the word level.
Further we investigated how individual words occur within recording sessions, focusing on recurrences within short time spans. We hypothesize that recurrence clusters — or "spikes" —sharpen children's attention for words in an interactional unit and assist them in learning. We found that spikes are universal in the diverse languages in the ACQDIV database. Moreover, different parts-of-speech display rather different spike characteristics. For instance, nouns tend to show stronger spikes than verbs. Interestingly, these differences replicate across all languages, suggesting a universal pattern.

Reported by

Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top