|
Learning complex center embedded structures with long distance dependencies, the easy way| old_uid | 11280 |
|---|
| title | Learning complex center embedded structures with long distance dependencies, the easy way |
|---|
| start_date | 2012/04/13 |
|---|
| schedule | 11h-12h |
|---|
| online | no |
|---|
| location_info | Amphi de Sciences Naturelles |
|---|
| summary | Lai & Poletiek (2011) found learning of a complex center embedded grammar (an AnBn grammar generating sentences of the type a1a2a3?b3b2b1, with long distance dependencies, e.g., a1(?.)b1) in an Artificial Grammar Learning task, under two necessary conditions of exposure: First the learning stimuli had to be presented in a growing fashion (first the most simple exemplars -without embeddings-, gradually followed by more complex ones). Second, when the simple items were omitted from the learning set, no learning occurred. Hence, for a center embedded structure to be learned from mere exposure to exemplars, it seems essential that exemplars without embeddings are learned thoroughly and in the earliest stage. The second effect is the focus of the present talk.
I will propose a statistical account of the effect of early simple adjacencies learning, and discuss how this effect could operate in natural grammar learning. |
|---|
| responsibles | Pélissier |
|---|
| |
|