DocumentCode :
288669
Title :
The English past tense-is this the extent of language learning by connectionist networks?
Author :
Daugherty, Kim G. ; Hare, Mary
Author_Institution :
Dept. of Comput. Sci., Univ. of Southern California, Los Angeles, CA, USA
Volume :
4
fYear :
1994
fDate :
27 Jun-2 Jul 1994
Firstpage :
2303
Abstract :
One potential problem for connectionist models of inflectional morphology is their ability to learn a “default” inflection. The early model of Rumelhart and McClelland (1986) appears to treat the regular English past tense (add -ed) as a default inflection; but Prasada and Pinker (1993) claim that this network can learn a default inflection only if it applies to a much larger class of items than any other inflection. The question of interest, however, is whether the claim is true only of that model, or whether it can be made of the connectionist approach in general. In the current paper we will show that the problem is not a general one, and that connectionist models are perfectly capable of learning a default category and generalizing as required, even without the benefit of superior class size
Keywords :
learning (artificial intelligence); mathematical morphology; natural languages; neural nets; English past tense; connectionist networks; inflectional morphology; natural language learning; Aircraft; Computer science; Educational institutions; Frequency; Morphology; Natural languages; Proposals; Psychology; Shape; Surface treatment;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Neural Networks, 1994. IEEE World Congress on Computational Intelligence., 1994 IEEE International Conference on
Conference_Location :
Orlando, FL
Print_ISBN :
0-7803-1901-X
Type :
conf
DOI :
10.1109/ICNN.1994.374578
Filename :
374578
Link To Document :
بازگشت