Improving Machine Translation Quality Prediction with Syntactic Tree Kernels

Christian Hardmeier

Improving Machine Translation Quality Prediction with Syntactic Tree Kernels

Christian Hardmeier

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

We investigate the problem of predicting the quality of a given Machine Translation (MT) output segment as a binary classification task. In a study with four different data sets in two text genres and two language pairs, we show that the performance of a Support Vector Machine (SVM) classifier can be improved by extending the feature set with implicitly defined syntactic features in the form of tree kernels over syntactic parse trees. Moreover, we demonstrate that syntax tree kernels achieve surprisingly high performance levels even without additional features, which makes them suitable as a low-effort initial building block for an MT quality estimation system.

Original language	English
Title of host publication	Proceedings of the 15th International Conference of the European Association for Machine Translation
Publisher	European Association for Machine Translation
Pages	233-240
Number of pages	8
Publication status	Published - 31 May 2011
Event	15th Annual Conference of the European Association for Machine Translation - Leuven, Belgium Duration: 30 May 2011 → 31 May 2011

Conference

Conference	15th Annual Conference of the European Association for Machine Translation
Country/Territory	Belgium
City	Leuven
Period	30/05/11 → 31/05/11

Access to Document

Improving Machine Translation_HARDMEIER_DOA22102019_VOR_CC-BY-NC-NDFinal published version, 632 KBLicence: Creative Commons: Attribution-NonCommercial-ShareAlike (CC BY-NC-SA)

http://www.mt-archive.info/10/EAMT-2011-TOC.htmLicence: Creative Commons: Attribution-NonCommercial-ShareAlike (CC BY-NC-SA)

Cite this

@inproceedings{88a2cbda637844368eba4da7f49b860f,

title = "Improving Machine Translation Quality Prediction with Syntactic Tree Kernels",

abstract = "We investigate the problem of predicting the quality of a given Machine Translation (MT) output segment as a binary classification task. In a study with four different data sets in two text genres and two language pairs, we show that the performance of a Support Vector Machine (SVM) classifier can be improved by extending the feature set with implicitly defined syntactic features in the form of tree kernels over syntactic parse trees. Moreover, we demonstrate that syntax tree kernels achieve surprisingly high performance levels even without additional features, which makes them suitable as a low-effort initial building block for an MT quality estimation system.",

author = "Christian Hardmeier",

year = "2011",

month = may,

day = "31",

language = "English",

pages = "233--240",

booktitle = "Proceedings of the 15th International Conference of the European Association for Machine Translation",

publisher = "European Association for Machine Translation",

note = "15th Annual Conference of the European Association for Machine Translation ; Conference date: 30-05-2011 Through 31-05-2011",

}

Hardmeier, C 2011, Improving Machine Translation Quality Prediction with Syntactic Tree Kernels. in Proceedings of the 15th International Conference of the European Association for Machine Translation. European Association for Machine Translation, pp. 233-240, 15th Annual Conference of the European Association for Machine Translation, Leuven, Belgium, 30/05/11. <http://www.mt-archive.info/10/EAMT-2011-TOC.htm>

TY - GEN

T1 - Improving Machine Translation Quality Prediction with Syntactic Tree Kernels

AU - Hardmeier, Christian

PY - 2011/5/31

Y1 - 2011/5/31

N2 - We investigate the problem of predicting the quality of a given Machine Translation (MT) output segment as a binary classification task. In a study with four different data sets in two text genres and two language pairs, we show that the performance of a Support Vector Machine (SVM) classifier can be improved by extending the feature set with implicitly defined syntactic features in the form of tree kernels over syntactic parse trees. Moreover, we demonstrate that syntax tree kernels achieve surprisingly high performance levels even without additional features, which makes them suitable as a low-effort initial building block for an MT quality estimation system.

AB - We investigate the problem of predicting the quality of a given Machine Translation (MT) output segment as a binary classification task. In a study with four different data sets in two text genres and two language pairs, we show that the performance of a Support Vector Machine (SVM) classifier can be improved by extending the feature set with implicitly defined syntactic features in the form of tree kernels over syntactic parse trees. Moreover, we demonstrate that syntax tree kernels achieve surprisingly high performance levels even without additional features, which makes them suitable as a low-effort initial building block for an MT quality estimation system.

M3 - Conference contribution

SP - 233

EP - 240

BT - Proceedings of the 15th International Conference of the European Association for Machine Translation

PB - European Association for Machine Translation

T2 - 15th Annual Conference of the European Association for Machine Translation

Y2 - 30 May 2011 through 31 May 2011

ER -

Improving Machine Translation Quality Prediction with Syntactic Tree Kernels

Abstract

Conference

Access to Document

Fingerprint

Cite this