Machine translation evaluation with neural networks

Francisco Guzmán, Shafiq Joty, Lluis Marques, Preslav Nakov

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

We present a framework for machine translation evaluation using neural networks in a pairwise setting, where the goal is to select the better translation from a pair of hypotheses, given the reference translation. In this framework, lexical, syntactic and semantic information from the reference and the two hypotheses is embedded into compact distributed vector representations, and fed into a multi-layer neural network that models nonlinear interactions between each of the hypotheses and the reference, as well as between the two hypotheses. We experiment with the benchmark datasets from the WMT Metrics shared task, on which we obtain the best results published so far, with the basic network configuration. We also perform a series of experiments to analyze and understand the contribution of the different components of the network. We evaluate variants and extensions, including fine-tuning of the semantic embeddings, and sentence-based representations modeled with convolutional and recurrent neural networks. In summary, the proposed framework is flexible and generalizable, allows for efficient learning and scoring, and provides an MT evaluation metric that correlates with human judgments, and is on par with the state of the art.

Original languageEnglish
JournalComputer Speech and Language
DOIs
Publication statusAccepted/In press - 1 Jun 2016

Fingerprint

Machine Translation
Semantics
Neural Networks
Neural networks
Recurrent neural networks
Evaluation
Multilayer neural networks
Syntactics
Tuning
Experiments
Multilayer Neural Network
Metric
Nonlinear Interaction
Recurrent Neural Networks
Scoring
Neural Network Model
Correlate
Experiment
Pairwise
Benchmark

Keywords

  • Deep neural networks
  • Distributed representation of texts
  • Machine translation
  • Reference-based MT evaluation
  • Textual similarity

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Software
  • Human-Computer Interaction

Cite this

Machine translation evaluation with neural networks. / Guzmán, Francisco; Joty, Shafiq; Marques, Lluis; Nakov, Preslav.

In: Computer Speech and Language, 01.06.2016.

Research output: Contribution to journalArticle

Guzmán, Francisco ; Joty, Shafiq ; Marques, Lluis ; Nakov, Preslav. / Machine translation evaluation with neural networks. In: Computer Speech and Language. 2016.
@article{fe7ffbb34dac437192bc2ddc909b7637,
title = "Machine translation evaluation with neural networks",
abstract = "We present a framework for machine translation evaluation using neural networks in a pairwise setting, where the goal is to select the better translation from a pair of hypotheses, given the reference translation. In this framework, lexical, syntactic and semantic information from the reference and the two hypotheses is embedded into compact distributed vector representations, and fed into a multi-layer neural network that models nonlinear interactions between each of the hypotheses and the reference, as well as between the two hypotheses. We experiment with the benchmark datasets from the WMT Metrics shared task, on which we obtain the best results published so far, with the basic network configuration. We also perform a series of experiments to analyze and understand the contribution of the different components of the network. We evaluate variants and extensions, including fine-tuning of the semantic embeddings, and sentence-based representations modeled with convolutional and recurrent neural networks. In summary, the proposed framework is flexible and generalizable, allows for efficient learning and scoring, and provides an MT evaluation metric that correlates with human judgments, and is on par with the state of the art.",
keywords = "Deep neural networks, Distributed representation of texts, Machine translation, Reference-based MT evaluation, Textual similarity",
author = "Francisco Guzm{\'a}n and Shafiq Joty and Lluis Marques and Preslav Nakov",
year = "2016",
month = "6",
day = "1",
doi = "10.1016/j.csl.2016.12.005",
language = "English",
journal = "Computer Speech and Language",
issn = "0885-2308",
publisher = "Academic Press Inc.",

}

TY - JOUR

T1 - Machine translation evaluation with neural networks

AU - Guzmán, Francisco

AU - Joty, Shafiq

AU - Marques, Lluis

AU - Nakov, Preslav

PY - 2016/6/1

Y1 - 2016/6/1

N2 - We present a framework for machine translation evaluation using neural networks in a pairwise setting, where the goal is to select the better translation from a pair of hypotheses, given the reference translation. In this framework, lexical, syntactic and semantic information from the reference and the two hypotheses is embedded into compact distributed vector representations, and fed into a multi-layer neural network that models nonlinear interactions between each of the hypotheses and the reference, as well as between the two hypotheses. We experiment with the benchmark datasets from the WMT Metrics shared task, on which we obtain the best results published so far, with the basic network configuration. We also perform a series of experiments to analyze and understand the contribution of the different components of the network. We evaluate variants and extensions, including fine-tuning of the semantic embeddings, and sentence-based representations modeled with convolutional and recurrent neural networks. In summary, the proposed framework is flexible and generalizable, allows for efficient learning and scoring, and provides an MT evaluation metric that correlates with human judgments, and is on par with the state of the art.

AB - We present a framework for machine translation evaluation using neural networks in a pairwise setting, where the goal is to select the better translation from a pair of hypotheses, given the reference translation. In this framework, lexical, syntactic and semantic information from the reference and the two hypotheses is embedded into compact distributed vector representations, and fed into a multi-layer neural network that models nonlinear interactions between each of the hypotheses and the reference, as well as between the two hypotheses. We experiment with the benchmark datasets from the WMT Metrics shared task, on which we obtain the best results published so far, with the basic network configuration. We also perform a series of experiments to analyze and understand the contribution of the different components of the network. We evaluate variants and extensions, including fine-tuning of the semantic embeddings, and sentence-based representations modeled with convolutional and recurrent neural networks. In summary, the proposed framework is flexible and generalizable, allows for efficient learning and scoring, and provides an MT evaluation metric that correlates with human judgments, and is on par with the state of the art.

KW - Deep neural networks

KW - Distributed representation of texts

KW - Machine translation

KW - Reference-based MT evaluation

KW - Textual similarity

UR - http://www.scopus.com/inward/record.url?scp=85011076232&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85011076232&partnerID=8YFLogxK

U2 - 10.1016/j.csl.2016.12.005

DO - 10.1016/j.csl.2016.12.005

M3 - Article

JO - Computer Speech and Language

JF - Computer Speech and Language

SN - 0885-2308

ER -