Modeling the Development of Lexicon with a Growing Self-Organizing Map

Igor Farkaš, Ping Li

Research output: Chapter in Book/Report/Conference proceedingConference contribution

9 Citations (Scopus)

Abstract

We present a self-organizing neural network model that can acquire an incremental lexicon. The model allows the acquisition of new words without disrupting learned structure. The model consists of three major components. First, the word co-occurrence detector computes word transition probabilities and represents word meanings in terms of context vectors. Second, word representations are projected to a lower, constant dimension. Third, the growing lexical map (GLM) self-organizes on the dimension-reduced word representations. The model is initialized with a subset of units in GLM and a subset of the lexicon, which enables it to capture the regularities of the input space and decrease chances of catastrophic interference. During growth, new nodes are inserted in order to reduce the map quantization error, and the insertion occurs only to yet unoccupied grid positions, thus preserving the 2D map topology. We have tested GLM on a portion of parental speech extracted from the CHILDES database, with an initial 200 words scattered among 800 nodes. The model demonstrates the ability to highly preserve learned lexical structure when 100 new words are gradually added. Implications of the model are discussed with respect to language acquisition by children.

Original languageEnglish (US)
Title of host publicationProceedings of the 6th Joint Conference on Information Sciences, JCIS 2002
EditorsJ.H. Caulfield, S.H. Chen, H.D. Cheng, R. Duro, J.H. Caufield, S.H. Chen, H.D. Cheng, R. Duro, V. Honavar
Pages553-556
Number of pages4
Volume6
StatePublished - 2002
EventProceedings of the 6th Joint Conference on Information Sciences, JCIS 2002 - Research Triange Park, NC, United States
Duration: Mar 8 2002Mar 13 2002

Other

OtherProceedings of the 6th Joint Conference on Information Sciences, JCIS 2002
CountryUnited States
CityResearch Triange Park, NC
Period3/8/023/13/02

Fingerprint

Self organizing maps
Topology
Detectors
Neural networks

All Science Journal Classification (ASJC) codes

  • Computer Science(all)

Cite this

Farkaš, I., & Li, P. (2002). Modeling the Development of Lexicon with a Growing Self-Organizing Map. In J. H. Caulfield, S. H. Chen, H. D. Cheng, R. Duro, J. H. Caufield, S. H. Chen, H. D. Cheng, R. Duro, ... V. Honavar (Eds.), Proceedings of the 6th Joint Conference on Information Sciences, JCIS 2002 (Vol. 6, pp. 553-556)
Farkaš, Igor ; Li, Ping. / Modeling the Development of Lexicon with a Growing Self-Organizing Map. Proceedings of the 6th Joint Conference on Information Sciences, JCIS 2002. editor / J.H. Caulfield ; S.H. Chen ; H.D. Cheng ; R. Duro ; J.H. Caufield ; S.H. Chen ; H.D. Cheng ; R. Duro ; V. Honavar. Vol. 6 2002. pp. 553-556
@inproceedings{ed04820e8fa54609a66331c0f09d8b46,
title = "Modeling the Development of Lexicon with a Growing Self-Organizing Map",
abstract = "We present a self-organizing neural network model that can acquire an incremental lexicon. The model allows the acquisition of new words without disrupting learned structure. The model consists of three major components. First, the word co-occurrence detector computes word transition probabilities and represents word meanings in terms of context vectors. Second, word representations are projected to a lower, constant dimension. Third, the growing lexical map (GLM) self-organizes on the dimension-reduced word representations. The model is initialized with a subset of units in GLM and a subset of the lexicon, which enables it to capture the regularities of the input space and decrease chances of catastrophic interference. During growth, new nodes are inserted in order to reduce the map quantization error, and the insertion occurs only to yet unoccupied grid positions, thus preserving the 2D map topology. We have tested GLM on a portion of parental speech extracted from the CHILDES database, with an initial 200 words scattered among 800 nodes. The model demonstrates the ability to highly preserve learned lexical structure when 100 new words are gradually added. Implications of the model are discussed with respect to language acquisition by children.",
author = "Igor Farkaš and Ping Li",
year = "2002",
language = "English (US)",
isbn = "0970789017",
volume = "6",
pages = "553--556",
editor = "J.H. Caulfield and S.H. Chen and H.D. Cheng and R. Duro and J.H. Caufield and S.H. Chen and H.D. Cheng and R. Duro and V. Honavar",
booktitle = "Proceedings of the 6th Joint Conference on Information Sciences, JCIS 2002",

}

Farkaš, I & Li, P 2002, Modeling the Development of Lexicon with a Growing Self-Organizing Map. in JH Caulfield, SH Chen, HD Cheng, R Duro, JH Caufield, SH Chen, HD Cheng, R Duro & V Honavar (eds), Proceedings of the 6th Joint Conference on Information Sciences, JCIS 2002. vol. 6, pp. 553-556, Proceedings of the 6th Joint Conference on Information Sciences, JCIS 2002, Research Triange Park, NC, United States, 3/8/02.

Modeling the Development of Lexicon with a Growing Self-Organizing Map. / Farkaš, Igor; Li, Ping.

Proceedings of the 6th Joint Conference on Information Sciences, JCIS 2002. ed. / J.H. Caulfield; S.H. Chen; H.D. Cheng; R. Duro; J.H. Caufield; S.H. Chen; H.D. Cheng; R. Duro; V. Honavar. Vol. 6 2002. p. 553-556.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Modeling the Development of Lexicon with a Growing Self-Organizing Map

AU - Farkaš, Igor

AU - Li, Ping

PY - 2002

Y1 - 2002

N2 - We present a self-organizing neural network model that can acquire an incremental lexicon. The model allows the acquisition of new words without disrupting learned structure. The model consists of three major components. First, the word co-occurrence detector computes word transition probabilities and represents word meanings in terms of context vectors. Second, word representations are projected to a lower, constant dimension. Third, the growing lexical map (GLM) self-organizes on the dimension-reduced word representations. The model is initialized with a subset of units in GLM and a subset of the lexicon, which enables it to capture the regularities of the input space and decrease chances of catastrophic interference. During growth, new nodes are inserted in order to reduce the map quantization error, and the insertion occurs only to yet unoccupied grid positions, thus preserving the 2D map topology. We have tested GLM on a portion of parental speech extracted from the CHILDES database, with an initial 200 words scattered among 800 nodes. The model demonstrates the ability to highly preserve learned lexical structure when 100 new words are gradually added. Implications of the model are discussed with respect to language acquisition by children.

AB - We present a self-organizing neural network model that can acquire an incremental lexicon. The model allows the acquisition of new words without disrupting learned structure. The model consists of three major components. First, the word co-occurrence detector computes word transition probabilities and represents word meanings in terms of context vectors. Second, word representations are projected to a lower, constant dimension. Third, the growing lexical map (GLM) self-organizes on the dimension-reduced word representations. The model is initialized with a subset of units in GLM and a subset of the lexicon, which enables it to capture the regularities of the input space and decrease chances of catastrophic interference. During growth, new nodes are inserted in order to reduce the map quantization error, and the insertion occurs only to yet unoccupied grid positions, thus preserving the 2D map topology. We have tested GLM on a portion of parental speech extracted from the CHILDES database, with an initial 200 words scattered among 800 nodes. The model demonstrates the ability to highly preserve learned lexical structure when 100 new words are gradually added. Implications of the model are discussed with respect to language acquisition by children.

UR - http://www.scopus.com/inward/record.url?scp=1642279513&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=1642279513&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:1642279513

SN - 0970789017

VL - 6

SP - 553

EP - 556

BT - Proceedings of the 6th Joint Conference on Information Sciences, JCIS 2002

A2 - Caulfield, J.H.

A2 - Chen, S.H.

A2 - Cheng, H.D.

A2 - Duro, R.

A2 - Caufield, J.H.

A2 - Chen, S.H.

A2 - Cheng, H.D.

A2 - Duro, R.

A2 - Honavar, V.

ER -

Farkaš I, Li P. Modeling the Development of Lexicon with a Growing Self-Organizing Map. In Caulfield JH, Chen SH, Cheng HD, Duro R, Caufield JH, Chen SH, Cheng HD, Duro R, Honavar V, editors, Proceedings of the 6th Joint Conference on Information Sciences, JCIS 2002. Vol. 6. 2002. p. 553-556