This chapter argues that claims for recurrent networks as plausible models of human sentence processing from which generalizations that have seemed to require the mediation of symbolically represented grammars are" emergent" are misplaced. Rather, it is argued that the linguistic relevance of connectionist networks lies in their application to lexicalist grammar induction. i. What do SRNs compute?
|Title of host publication||The Lexical Basis of Sentence Processing|
|Subtitle of host publication||Formal, computational and experimental issues|
|Editors||Paola Merlo, Suzanne Stevenson|
|Publisher||John Benjamins Publishing Company|
|Number of pages||14|
|Publication status||Published - Aug 2002|