Improved Statistical Machine Translation Using Paraphrases

By Chris Callison-Burch, Philipp Koehn, Miles Osborne. North American Chapter of the Association for Computational Linguistics, 2006

338 citations recorded by the source corpus.

Parallel corpora are crucial for training SMT systems. However, for many language pairs they are available only in very limited quantities. For these language pairs a huge portion of phrases encountered at run-time will be unknown. We show how techniques from paraphrasing can be used to deal with these otherwise unknown source language phrases. Our results show that augmenting a state-of-the-art SMT system with paraphrases leads to significantly improved coverage and translation quality. For a training corpus with 10,000 sentence pairs we increase the coverage of unique test set unigrams from 48% to 90%, with more than half of the newly covered items accurately translated, as opposed to none in current approaches.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/b4491aa9c5a906f060b5c9b6b24db3b11cd44238

How researchers see this paper

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

In Artificial Intelligence

Citations in the corpus

Simple and Effective Paraphrastic Similarity from Parallel Translations (2019, background): “Examples include bilingual pivoting over phrases (Callison-Burch et al., 2006; Ganitkevitch et al., 2013), and automatic translation of one side of the bitext (Wieting et al.”
A Comparison of Pivot Methods for Phrase-Based Statistical Machine Translation (2007, background): “Callison-Burch et al. (2006a) use a similar deﬁnition for a paraphrase probability. where count ( f, e ) gives the total number of times the word f is aligned with the word e in the parallel corpus.”
Machine Translation by Triangulation: Making Effective Use of Multi-Parallel Corpora (2007, methodology): “Although related to Callison-Burch et al. (2006) our method is conceptually simpler and more general.”
Paraphrasing Revisited with Neural Machine Translation (2017, background): “Examples include query and pattern expansion (Riezler et al., 2007), summarization (Barzilay, 2003), question answering (Lin and Pantel, 2001), semantic parsing (Berant and Liang, 2014), semantic role labeling (Woodsend and La-pata, 2014), and machine translation (Callison-Burch et al., 2006).”
Toward Robust Neural Machine Translation for Noisy Input Sequences (2017, methodology): “Inducing noise in the training inputs can be seen as a form of data augmentation, which has been used in several applications such as acoustic modeling [16], computer vision [17], language modeling [18], and statistical machine translation where data can be augmented by paraphrases [19].”
Collecting Highly Parallel Data for Paraphrase Evaluation (2011, background): “…judges (Barzilay and McKeown, 2001; Ibrahim et al., 2003; Bannard and Callison-Burch, 2005) or indirect, task-based methods (Lin and Pantel, 2001; Callison-Burch et al., 2006), there have also been a few attempts at creating automatic metrics that can be more easily replicated and used to…”
Simple is Better! Lightweight Data Augmentation for Low Resource Slot Filling and Intent Classification (2020, background): “Yang, 2015; Kobayashi, 2018; Wei and Zou, 2019), to manipulation of the sentence structure (Şahin and Steedman, 2018b), to paraphrase-based augmentation (Callison-Burch et al., 2006).”
Enriching Parallel Corpora for Statistical Machine Translation with Semantic Negation Rephrasing (2012, background): “…paraphrasing data in order to overcome the limitations that insufﬁ-ciently large or underrepresented phenomena in par-1 allel corpora impose on SMT. Callison-Burch et al. (2006) tackle the problem of unseen phrases in SMT by adding source language paraphrases to the phrase table with appropriate…”
Pivot language approach for phrase-based statistical machine translation (2007, methodology): “Callison-Burch et al. (2006) used pivot languages for paraphrase extraction to handle the unseen phrases for phrase-based SMT. Borin (2000) and Wang et al. (2006) used pivot languages to improve word alignment.”
Improving machine translation via triangulation and transliteration (2014, positive): “Techniques like triangulation (Cohn and Lapata, 2007; Wu and Wang, 2007) and paraphrasing (Callison-Burch et al., 2006) have also been used to address the problem of data sparsity.”
Paraphrase Identification as Probabilistic Quasi-Synchronous Recognition (2009, background): “For natural language engineers, the problem bears on information management systems like abstractive summarizers that must measure semantic overlap between sentences (Barzi-lay and Lee, 2003), question answering modules (Marsi and Krahmer, 2005) and machine translation (Callison-Burch et al., 2006).”
IXA Biomedical Translation System at WMT16 Biomedical Translation Task (2016, background): “With respect to transliteration, (Callison-Burch et al., 2006) exposes that state-of-the-art systems usually apply two strategies to cope with OOV words, neither of them satisfactory.”
Improving Arabic-Chinese Statistical Machine Translation using English as Pivot Language (2009, background): “Callison-Burch et al (2006) exploited the existence of multiple parallel corpora to learn paraphrases for Phrase-based MT. Filali and Bilmes (2005) improved word alignment by leveraging multilingual parallel translations.”
Paraphrase Augmented Task-Oriented Dialog Generation (2020, methodology): “augmentation methods have been proved to be useful in various tasks, such as machine translation (Callison-Burch et al., 2006), text classification (Zhang et al.”
Large Scale Acquisition of Paraphrases for Learning Surface Patterns (2008, methodology): “For example, in question answering, paraphrases have been used to find multiple patterns that pinpoint the same answer (Ravichandran and Hovy, 2002); in statistical machine translation, they have been used to find translations for unseen source language phrases (Callison-Burch et al., 2006); in mult”
Combining Word-Level and Character-Level Models for Machine Translation Between Closely-Related Languages (2012, methodology): “Then, following (Callison-Burch et al., 2006; Wu and Wang, 2007; Utiyama and Isahara, 2007), we induced conditional lexical translation probabilities as Pr( m | b ) = e Pr( m | e ) Pr( e | b ) , where Pr( m | e ) and Pr( e | b ) are estimated using maximum likelihood from MK–EN and EN–BG word…”