A new machine translation evaluation metric utilizing the holder mean

Rabha W. Ibrahim, Arwa Hatem, Nazlia Omar

Research output: Contribution to journalArticle

Abstract

It is significant to recognize that user require the best translation system; when their requirement is to translate text from one language to another. Human assessment of Machine Translation (MT) production takes a long time to finish and includes a great deal of effort that cannot be reprocessed. Many automatic assessment metrics have been presented to overcome some of the drawbacks of MT evaluation i.e., BLEU, which only measures word reordering circuitously. This paper presents a new statistical method, based on a comparison between machine translations and reference sentences, which will be utilized as a direct method for measuring word order. This method reflects the organization of machine translation sentences. We will show that our statistical method associates very well with human judgments and that this method has numerous beneficial properties i.e., it gives a fitting of a dataset; in that it instructsdirect this data successively (without losing generality). Furthermore, this method calculates values founded by all texts. It depends on the simplicity of the statistical analysis method which is not affected by varying samples. it practices algebraically; and it brings and represents all data properties. The evaluation method presented here was applied to Arabic-English machine translation. Our method is essentially defined by the fractional Holder mean. The assessment outcomes of the new method indicate that the statistical method used works well with the BLEU method in evaluating machine translation systems.

Original languageEnglish
Pages (from-to)1421-1434
Number of pages14
JournalPakistan Journal of Statistics
Volume30
Issue number6
Publication statusPublished - 1 Jan 2014

Fingerprint

Machine Translation
Metric
Evaluation
Statistical method
Reordering
Evaluation Method
Direct Method
Statistical Analysis
Simplicity
Fractional
Calculate
Requirements

Keywords

  • BLEU
  • Holder mean
  • Machine translation
  • Statistical analysis

ASJC Scopus subject areas

  • Statistics and Probability

Cite this

A new machine translation evaluation metric utilizing the holder mean. / Ibrahim, Rabha W.; Hatem, Arwa; Omar, Nazlia.

In: Pakistan Journal of Statistics, Vol. 30, No. 6, 01.01.2014, p. 1421-1434.

Research output: Contribution to journalArticle

Ibrahim, Rabha W. ; Hatem, Arwa ; Omar, Nazlia. / A new machine translation evaluation metric utilizing the holder mean. In: Pakistan Journal of Statistics. 2014 ; Vol. 30, No. 6. pp. 1421-1434.
@article{ab3a1ffec2dc43ef98e19a0d9eb67a32,
title = "A new machine translation evaluation metric utilizing the holder mean",
abstract = "It is significant to recognize that user require the best translation system; when their requirement is to translate text from one language to another. Human assessment of Machine Translation (MT) production takes a long time to finish and includes a great deal of effort that cannot be reprocessed. Many automatic assessment metrics have been presented to overcome some of the drawbacks of MT evaluation i.e., BLEU, which only measures word reordering circuitously. This paper presents a new statistical method, based on a comparison between machine translations and reference sentences, which will be utilized as a direct method for measuring word order. This method reflects the organization of machine translation sentences. We will show that our statistical method associates very well with human judgments and that this method has numerous beneficial properties i.e., it gives a fitting of a dataset; in that it instructsdirect this data successively (without losing generality). Furthermore, this method calculates values founded by all texts. It depends on the simplicity of the statistical analysis method which is not affected by varying samples. it practices algebraically; and it brings and represents all data properties. The evaluation method presented here was applied to Arabic-English machine translation. Our method is essentially defined by the fractional Holder mean. The assessment outcomes of the new method indicate that the statistical method used works well with the BLEU method in evaluating machine translation systems.",
keywords = "BLEU, Holder mean, Machine translation, Statistical analysis",
author = "Ibrahim, {Rabha W.} and Arwa Hatem and Nazlia Omar",
year = "2014",
month = "1",
day = "1",
language = "English",
volume = "30",
pages = "1421--1434",
journal = "Pakistan Journal of Statistics",
issn = "1012-9367",
publisher = "Pakistan Journal of Statistics",
number = "6",

}

TY - JOUR

T1 - A new machine translation evaluation metric utilizing the holder mean

AU - Ibrahim, Rabha W.

AU - Hatem, Arwa

AU - Omar, Nazlia

PY - 2014/1/1

Y1 - 2014/1/1

N2 - It is significant to recognize that user require the best translation system; when their requirement is to translate text from one language to another. Human assessment of Machine Translation (MT) production takes a long time to finish and includes a great deal of effort that cannot be reprocessed. Many automatic assessment metrics have been presented to overcome some of the drawbacks of MT evaluation i.e., BLEU, which only measures word reordering circuitously. This paper presents a new statistical method, based on a comparison between machine translations and reference sentences, which will be utilized as a direct method for measuring word order. This method reflects the organization of machine translation sentences. We will show that our statistical method associates very well with human judgments and that this method has numerous beneficial properties i.e., it gives a fitting of a dataset; in that it instructsdirect this data successively (without losing generality). Furthermore, this method calculates values founded by all texts. It depends on the simplicity of the statistical analysis method which is not affected by varying samples. it practices algebraically; and it brings and represents all data properties. The evaluation method presented here was applied to Arabic-English machine translation. Our method is essentially defined by the fractional Holder mean. The assessment outcomes of the new method indicate that the statistical method used works well with the BLEU method in evaluating machine translation systems.

AB - It is significant to recognize that user require the best translation system; when their requirement is to translate text from one language to another. Human assessment of Machine Translation (MT) production takes a long time to finish and includes a great deal of effort that cannot be reprocessed. Many automatic assessment metrics have been presented to overcome some of the drawbacks of MT evaluation i.e., BLEU, which only measures word reordering circuitously. This paper presents a new statistical method, based on a comparison between machine translations and reference sentences, which will be utilized as a direct method for measuring word order. This method reflects the organization of machine translation sentences. We will show that our statistical method associates very well with human judgments and that this method has numerous beneficial properties i.e., it gives a fitting of a dataset; in that it instructsdirect this data successively (without losing generality). Furthermore, this method calculates values founded by all texts. It depends on the simplicity of the statistical analysis method which is not affected by varying samples. it practices algebraically; and it brings and represents all data properties. The evaluation method presented here was applied to Arabic-English machine translation. Our method is essentially defined by the fractional Holder mean. The assessment outcomes of the new method indicate that the statistical method used works well with the BLEU method in evaluating machine translation systems.

KW - BLEU

KW - Holder mean

KW - Machine translation

KW - Statistical analysis

UR - http://www.scopus.com/inward/record.url?scp=85065234002&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85065234002&partnerID=8YFLogxK

M3 - Article

VL - 30

SP - 1421

EP - 1434

JO - Pakistan Journal of Statistics

JF - Pakistan Journal of Statistics

SN - 1012-9367

IS - 6

ER -