Improved Statistical Alignment Models

By Franz Joseph Och, Hermann Ney. Annual Meeting of the Association for Computational Linguistics, 2000

1,193 citations recorded by the source corpus.

In this paper, we present and compare various single-word based alignment models for statistical machine translation. We discuss the five IBM alignment models, the Hidden-Markov alignment model, smoothing techniques and various modifications. We present different methods to combine alignments. As evaluation criterion we use the quality of the resulting Viterbi alignment compared to a manually produced reference alignment. We show that models with a first-order dependence and a fertility model lead to significantly better results than the simple models IBM-1 or IBM-2, which are not able to go beyond zero-order dependencies.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/c9214ebe91454e6369720136ab7dd990d52a07d4

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

Chimera – Three Heads for English-to-Czech Translation (2013, methodology): “GIZA++ (Och and Ney, 2000) has been the standard tool for computing word alignment in phrase-based MT.”
Story Generation from Sequence of Independent Short Descriptions (2017, methodology): “GIZA++ is congured to apply the principles of IBM model 4 and 5 [6] and the HMM alignment model [23] for alignment learning.”
Improving Translation Model by Monolingual Data (2011, methodology): “, 2007) decoder and tools, SRILM (Stolcke, 2002) and KenLM (Heafield, 2011) for language modelling and GIZA++ (Och and Ney, 2000) for word alignments.”
Word Sense Disambiguation Improves Statistical Machine Translation (2007, methodology): “First, we performed word alignment on the FBIS parallel corpus using GIZA++ (Och and Ney, 2000) in both directions.”
Intent transfer in speech-to-speech machine translation (2012, methodology): “We use GIZA++ [12] tool to align the sentences within each language pair.”
Paraphrase-Driven Learning for Open Question Answering (2013, background): “Lastly, we believe that it would be possible to leverage the user-provided answers from WikiAnswers as a source of supervision.”
Semantic Parsing as Machine Translation (2013, methodology): “We also specify a decoding procedure that will return structured MRs for an utterance during prediction.”
Jointly Learning to Align and Translate with Transformer Models (2019, methodology): “The alignment quality is evaluated by using the alignment error rate (AER) introduced in (Och and Ney, 2000).”
Tree-to-String Alignment Template for Statistical Machine Translation (2006, methodology): “…we used for comparison was Pharaoh (Koehn et al., 2003; Koehn, 2004), a freely available decoder for phrase-based translation models: We ran GIZA++ (Och and Ney, 2000) on the training corpus in both directions using its default setting, and then applied the reﬁnement rule “diag-and” described in…”
Posterior Regularization for Structured Latent Variable Models (2010, methodology): “5 Results We evaluated the constraints using the Hansard corpus (Och and Ney, 2000) of English/French.”
Modeling Coherence for Neural Machine Translation with Dynamic and Topic Caches (2017, methodology): “We ran GIZA++ (Och and Ney, 2000) on the training data in both directions, and merged alignments in two directions with “grow-diag-final” refinement rule (Koehn et al.”
It Makes Sense: A Wide-Coverage Word Sense Disambiguation System for Free Text (2010, methodology): “The framework of IMS allows us to integrate different preprocessing tools to generate knowledge sources.”
Neural Machine Translation with Supervised Attention (2016, background): “However, it differs from conventional alignment models that are able to use the target word to infer its alignments (Och and Ney, 2000; Dyer et al., 2013; Liu and Sun, 2015), and as a result there is a substantial gap in quality between the alignments derived by this attention based NMT and…”
Minimum Bayes-Risk Word Alignments of Bilingual Texts (2002, methodology): “Our unseen test data consisted of 207 FrenchEnglish sentence pairs from the Hansards corpus (Och and Ney, 2000b).”
The impact of parse quality on syntactically-informed statistical machine translation (2006, methodology): “First the corpus is word-aligned using GIZA++ (Och and Ney, 2000), then the source sentence are parsed, and ﬁnally dependencies are projected onto the target side following the heuristics described in (Quirk et al., 2005).”