Performance of a SCFG-based language model with training data sets of increasing size

Joan Andreu Sánchez; José Miguel Benedí; Diego Linares

doi:10.1007/11492542_72

Performance of a SCFG-based language model with training data sets of increasing size

Joan Andreu Sánchez, José Miguel Benedí, Diego Linares

Polytechnic University of Valencia

Producción: Contribución a una revista › Artículo de la conferencia › revisión exhaustiva

1 Cita (Scopus)

Resumen

In this paper, a hybrid language model which combines a word-based n-gram and a category-based Stochastic Context-Free Grammar (SCFG) is evaluated for training data sets of increasing size. Different estimation algorithms for learning SCFGs in General Format and in Chomsky Normal Form are considered. Experiments on the UPenn Treebank corpus are reported. These experiments have been carried out in terms of the test set perplexity and the word error rate in a speech recognition experiment.

Idioma original	Inglés
Páginas (desde-hasta)	586-594
Número de páginas	9
Publicación	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volumen	3523
N.º	II
DOI	https://doi.org/10.1007/11492542_72
Estado	Publicada - 2005
Evento	Second Iberian Conference on Pattern Recognition and Image Analysis, IbPRIA 2005 - Estoril, Portugal Duración: 07 jun. 2005 → 09 jun. 2005

Acceder al documento

10.1007/11492542_72

Otros archivos y enlaces

Enlace a la publicación en Scopus

Citar esto

@article{e8c035afa2c74fec9b8377f6c2717e4f,

title = "Performance of a SCFG-based language model with training data sets of increasing size",

abstract = "In this paper, a hybrid language model which combines a word-based n-gram and a category-based Stochastic Context-Free Grammar (SCFG) is evaluated for training data sets of increasing size. Different estimation algorithms for learning SCFGs in General Format and in Chomsky Normal Form are considered. Experiments on the UPenn Treebank corpus are reported. These experiments have been carried out in terms of the test set perplexity and the word error rate in a speech recognition experiment.",

author = "S{\'a}nchez, {Joan Andreu} and Bened{\'i}, {Jos{\'e} Miguel} and Diego Linares",

year = "2005",

doi = "10.1007/11492542_72",

language = "English",

volume = "3523",

pages = "586--594",

journal = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

issn = "0302-9743",

publisher = "Springer Verlag",

number = "II",

note = "Second Iberian Conference on Pattern Recognition and Image Analysis, IbPRIA 2005 ; Conference date: 07-06-2005 Through 09-06-2005",

}

Performance of a SCFG-based language model with training data sets of increasing size. / Sánchez, Joan Andreu; Benedí, José Miguel; Linares, Diego.
En: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 3523, N.º II, 2005, p. 586-594.

Producción: Contribución a una revista › Artículo de la conferencia › revisión exhaustiva

TY - JOUR

T1 - Performance of a SCFG-based language model with training data sets of increasing size

AU - Sánchez, Joan Andreu

AU - Benedí, José Miguel

AU - Linares, Diego

PY - 2005

Y1 - 2005

N2 - In this paper, a hybrid language model which combines a word-based n-gram and a category-based Stochastic Context-Free Grammar (SCFG) is evaluated for training data sets of increasing size. Different estimation algorithms for learning SCFGs in General Format and in Chomsky Normal Form are considered. Experiments on the UPenn Treebank corpus are reported. These experiments have been carried out in terms of the test set perplexity and the word error rate in a speech recognition experiment.

AB - In this paper, a hybrid language model which combines a word-based n-gram and a category-based Stochastic Context-Free Grammar (SCFG) is evaluated for training data sets of increasing size. Different estimation algorithms for learning SCFGs in General Format and in Chomsky Normal Form are considered. Experiments on the UPenn Treebank corpus are reported. These experiments have been carried out in terms of the test set perplexity and the word error rate in a speech recognition experiment.

UR - http://www.scopus.com/inward/record.url?scp=25144434854&partnerID=8YFLogxK

U2 - 10.1007/11492542_72

DO - 10.1007/11492542_72

M3 - Conference article

AN - SCOPUS:25144434854

SN - 0302-9743

VL - 3523

SP - 586

EP - 594

JO - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

JF - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

IS - II

T2 - Second Iberian Conference on Pattern Recognition and Image Analysis, IbPRIA 2005

Y2 - 7 June 2005 through 9 June 2005

ER -

Performance of a SCFG-based language model with training data sets of increasing size

Resumen

Acceder al documento

Otros archivos y enlaces

Huella

Citar esto