An Investigation on the Effectiveness of Features for Translation Quality Estimation

By Kashif Shah, Trevor Cohn, Lucia Specia. 2013

48 citations recorded by the source corpus.

Subfields: Natural Language Processing, Machine Translation

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

In Machine Translation

Citations in the corpus

SHEF-Lite: When Less is More for Translation Quality Estimation (2013, methodology): “For task 1.3, we used a subset of the 80 most general BB features as in (Shah et al., 2013), for which we had all the necessary resources available for the extraction.”
SHEF-NN: Translation Quality Estimation with Neural Networks (2015, methodology): “• FS(AF) : Top 20 features selected from the above 82 features with Gaussian Processes (GPs) by the procedure described in (Shah et al., 2013b).”
SHEF-Lite 2.0: Sparse Multi-task Gaussian Processes for Translation Quality Estimation (2014, methodology): “While a variety of kernel functions are available, here we followed previous work in QE using GP (Cohn and Specia, 2013; Shah et al., 2013) and employed a squared exponential (SE) kernel with automatic relevance determination (ARD):”
Phrase-level Quality Estimation for Machine Translation (2015, background): “• LM and n-gram count features were extracted using LMs trained on the Europarl corpus using SRILM5.”
SHEF-LIUM-NN: Sentence level Quality Estimation with Neural Network Features (2016, methodology): “We extracted the following features: • QuEst : 79 black-box features using the QuEst framework (Specia et al., 2013; Shah et al., 2013a) as described in Shah et al. (2013b).”
Pushing the Limits of Translation Quality Estimation (2017, background): “Past approaches to this problem include linear classiﬁers with handcrafted features (Uefﬁng and Ney, 2007; Bic¸ici, 2013; Shah et al., 2013; Luong et al., 2014), sometimes subject to feature selection (Avramidis, 2012; Beck et al., 2013), recurrent neural networks (de Souza et al., 2014; Kim and…”
Findings of the 2013 Workshop on Statistical Machine Translation (2013, methodology): “To perform feature selection, following (Shah et al., 2013) features are ranked by the Gaussian Process algorithm according to their learned length scales, which can be interpreted as the relevance of such feature for the model.”
LORIA System for the WMT15 Quality Estimation Shared Task (2015, methodology): “( Scarton and Spe-cia, 2014) cite different works (Soricut et al., 2012; Shah et al., 2013) using also these measures for Quality Estimation.”
A temporal model of text periodicities using Gaussian Processes (2013, methodology): “(2011) and for Quality Estimation regression in (Cohn and Specia, 2013; Shah et al., 2013).”
Document-level translation quality estimation: exploring discourse and pseudo-references (2014, background): “1 Shah et al. (2013) conduct a feature analysis, at sentence-level, on a number of datasets and show that the BLEU-based pseudo-reference feature contributes the most to prediction performance.”
Metrics for Evaluation of Word-level Machine Translation Quality Estimation (2016, background): “…estimation (i.e. estimation of how conﬁdent a particular MT system is on a word or a phrase (Gan-drabur and Foster, 2003)) it evolved to system-independent QE and is performed at the word level (Luong et al., 2014), sentence level (Shah et al., 2013) and document level (Scarton et al., 2015).”
Exploring Consensus in Machine Translation for Quality Estimation (2014, methodology): “1 Shah et al. (2013) use pseudo-references in the same way to extract a BLEU feature for sentence-level prediction.”