Performance analysis of FFNN-based language model in contrast with n-gram

Kwang Ho Kim; Donghyun Lee; Minkyu Lim; Minho Ryang; Gil Jin Jang; Jeong Sik Park; Ji Hwan Kim

doi:10.1007/978-3-319-19291-8_25

Performance analysis of FFNN-based language model in contrast with n-gram

Kwang Ho Kim
, Donghyun Lee
, Minkyu Lim
, Minho Ryang
, Gil Jin Jang
, Jeong Sik Park
, Ji Hwan Kim

Sogang University

Research output: Chapter in Book/Report/Conference proceeding › Chapter › peer-review

Abstract

In this paper, we analyze the performance of feed forward neural network (FFNN)-based language model in contrast with n-gram. The probability of n-gram language model was estimated based on the statistics of word sequences. The FFNN-based language model was structured by three hidden layers, 500 hidden units per each hidden layer, and 30 dimension word embedding. The performance of FFNN-based language model is better than that of n-gram by 1.5 % in terms of WER on the English WSJ domain.

Original language	English
Title of host publication	Natural Language Dialog Systems and Intelligent Assistants
Publisher	Springer International Publishing
Pages	253-256
Number of pages	4
ISBN (Electronic)	9783319192918
ISBN (Print)	9783319192901
DOIs	https://doi.org/10.1007/978-3-319-19291-8_25
State	Published - 29 Oct 2015

Keywords

Feed forward neural network
Language model
N-Gram
Performance analysis

Access to Document

10.1007/978-3-319-19291-8_25

Cite this

@inbook{352a5554487a4c7ca4336bfbc25e2ae1,

title = "Performance analysis of FFNN-based language model in contrast with n-gram",

abstract = "In this paper, we analyze the performance of feed forward neural network (FFNN)-based language model in contrast with n-gram. The probability of n-gram language model was estimated based on the statistics of word sequences. The FFNN-based language model was structured by three hidden layers, 500 hidden units per each hidden layer, and 30 dimension word embedding. The performance of FFNN-based language model is better than that of n-gram by 1.5 \% in terms of WER on the English WSJ domain.",

keywords = "Feed forward neural network, Language model, N-Gram, Performance analysis",

author = "Kim, \{Kwang Ho\} and Donghyun Lee and Minkyu Lim and Minho Ryang and Jang, \{Gil Jin\} and Park, \{Jeong Sik\} and Kim, \{Ji Hwan\}",

note = "Publisher Copyright: {\textcopyright} Springer International Publishing Switzerland 2015. All rights are reserved.",

year = "2015",

month = oct,

day = "29",

doi = "10.1007/978-3-319-19291-8\_25",

language = "English",

isbn = "9783319192901",

pages = "253--256",

booktitle = "Natural Language Dialog Systems and Intelligent Assistants",

publisher = "Springer International Publishing",

address = "Switzerland",

}

TY - CHAP

T1 - Performance analysis of FFNN-based language model in contrast with n-gram

AU - Kim, Kwang Ho

AU - Lee, Donghyun

AU - Lim, Minkyu

AU - Ryang, Minho

AU - Jang, Gil Jin

AU - Park, Jeong Sik

AU - Kim, Ji Hwan

N1 - Publisher Copyright: © Springer International Publishing Switzerland 2015. All rights are reserved.

PY - 2015/10/29

Y1 - 2015/10/29

N2 - In this paper, we analyze the performance of feed forward neural network (FFNN)-based language model in contrast with n-gram. The probability of n-gram language model was estimated based on the statistics of word sequences. The FFNN-based language model was structured by three hidden layers, 500 hidden units per each hidden layer, and 30 dimension word embedding. The performance of FFNN-based language model is better than that of n-gram by 1.5 % in terms of WER on the English WSJ domain.

AB - In this paper, we analyze the performance of feed forward neural network (FFNN)-based language model in contrast with n-gram. The probability of n-gram language model was estimated based on the statistics of word sequences. The FFNN-based language model was structured by three hidden layers, 500 hidden units per each hidden layer, and 30 dimension word embedding. The performance of FFNN-based language model is better than that of n-gram by 1.5 % in terms of WER on the English WSJ domain.

KW - Feed forward neural network

KW - Language model

KW - N-Gram

KW - Performance analysis

UR - https://www.scopus.com/pages/publications/84955701476

U2 - 10.1007/978-3-319-19291-8_25

DO - 10.1007/978-3-319-19291-8_25

M3 - Chapter

AN - SCOPUS:84955701476

SN - 9783319192901

SP - 253

EP - 256

BT - Natural Language Dialog Systems and Intelligent Assistants

PB - Springer International Publishing

ER -

Performance analysis of FFNN-based language model in contrast with n-gram

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this