Informant learning is one of the approaches to language learning that was formalized by Gold (1967). It is a description of the language environment. In informant learning, a learner is presented both grammatical and ungrammatical examples of the language-to-be learned, and is also informed about the grammaticality of each example. In psychological terms, this is equivalent to a human child receiving both positive and negative evidence. Gold proved that complex grammars can only be identified in the limit via informant learning. This led to Gold’s paradox, because cognitive scientists later showed that children appear to learn complex languages with a much less powerful environment, which Gold called text learning.
Informant learning is also of interest because the supervised learning of connectionist networks can be thought of as an example of this sort of learning (in a domain that is usually nonlinguistic).
- Gold, E. M. (1967). Language identification in the limit. Information and Control, 10, 447-474.
(Added November, 2009)