Using Sense-labeled Discourse Connectives for Statistical Machine Translation

Thomas Meyer, Andrei Popescu-Belis
2012
Conference paper

This article shows how the automatic disambiguation of discourse connectives can improve Statistical Machine Translation (SMT) from English to French. Connectives are firstly disambiguated in terms of the discourse relation they signal between segments. Several classifiers trained using syntactic and semantic features reach state-of-the-art performance, with F1 scores of 0.6 to 0.8 over thirteen ambiguous English connectives. Labeled connectives are then used into SMT systems either by modifying their phrase table, or by training them on labeled corpora. The best modified SMT systems improve the translation of connectives without degrading BLEU scores. A threshold-based SMT system using only high-confidence labels improves BLEU scores by 0.2–0.4 points.

About this result

This page is automatically generated and may contain information that is not correct, complete, up-to-date, or relevant to your search query. The same applies to every other page on this website. Please make sure to verify the information with EPFL's official sources.

Related concepts

Related publications

Thomas Meyer, Andrei Popescu-Belis
2012
Conference paper

Abstract

Official source

https://infoscience.epfl.ch/record/192529?ln=en

About this result

Related concepts (6)

Related concepts

Related publications (16)

Related publications

Machine translation

Machine translation is use of either rule-based or probabilistic (i.e. statistical and, most recently, neural network-based) machine learning approaches to translation of text or speech from one la

Translation

Translation is the communication of the meaning of a source-language text by means of an equivalent target-language text. The English language draws a terminological distinction (which does not exis

Statistical machine translation

Statistical machine translation (SMT) was a machine translation approach, that superseded the previous, rule-based approach because it required explicit description of each and every linguistic rule,

Machine Translation with Many Manually Labeled Discourse Connectives

Thomas Meyer

The paper presents machine translation experiments from English to Czech with a large amount of manually annotated discourse connectives. The gold-standard discourse relation annotation leads to better translation performance in ranges of 4–60% for some ambiguous English connectives and helps to find correct syntactical constructs in Czech for less ambiguous connectives. Automatic scoring confirms the stability of the newly built discourse-aware translation systems. Error analysis and human translation evaluation point to the cases where the annotation was most and where less helpful.

2013

A Corpus-based Contrastive Analysis for Defining Minimal Semantics of Inter-sentential Dependencies for Machine Translation

Thomas Meyer, Andrei Popescu-Belis

Inter-sentential dependencies such as discourse connectives or pronouns have an impact on the translation of these items. These dependencies have classically been analyzed within complex theoretical frameworks, often monolingual ones, and the resulting fine-grained descriptions, although relevant to translation, are likely beyond reach of statistical machine translation systems. Instead, we propose an approach to search for a minimal, feature-based characterization of translation divergencies due to inter-sentential dependencies, in the case of discourse connectives and pronouns, based on contrastive analyses performed on the Europarl corpus. In addition, we show how to automatically assign labels to connectives and pronouns, and how to use them for statistical machine translation.

2011

Disambiguating Discourse Connectives for Statistical Machine Translation

Thomas Meyer, Andrei Popescu-Belis

This paper shows that the automatic labeling of discourse connectives with the relations they signal, prior to machine translation (MT), can be used by phrase-based statistical MT systems to improve their translations. This improvement is demonstrated here when translating from English to four target languages - French, German, Italian and Arabic - using several test sets from recent MT evaluation campaigns. Using automatically labeled data for training, tuning and testing MT systems is beneficial on condition that labels are sufficiently accurate, typically above 70%. To reach such an accuracy, a large array of features for discourse connective labeling (morpho-syntactic, semantic and discursive) are extracted using state-of-the-art tools and exploited in factored MT models. The translation of connectives is improved significantly, between 0.7% and 10% as measured with the dedicated ACT metric. The improvements depend mainly on the level of ambiguity of the connectives in the test sets.

2015