Forest Rescoring: Faster Decoding with Integrated Language Models

By Liang Huang, David Chiang. 2007

331 citations recorded by the source corpus.

Subfields: Natural Language Processing, Model Architectures

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 Model Architectures

Citations in the corpus

Edinburgh’s Phrase-based Machine Translation Systems for WMT-14 (2014, background): “We trained our systems with the following settings: a maximum sentence length of 80, grow-diag-ﬁnal-and symmetrization of GIZA++ alignments, an interpolated Kneser-Ney smoothed 5-gram language model with KenLM (Heaﬁeld, 2011) used at runtime, hierarchical lexicalized re-ordering (Galley and Manning,”
A Challenge Set Approach to Evaluating Machine Translation (2017, methodology): “Decoding used the cube-pruning algorithm of Huang and Chiang (2007), with a distortion limit of 7.”
EU-BRIDGE MT: Combined Machine Translation (2014, background): “, 2012), a distortion limit of 6, a maximum phrase length of 5, 100-best translation options, Minimum Bayes Risk decoding (Kumar and Byrne, 2004), cube pruning (Huang and Chiang, 2007), with a stack size of 1000 during tuning and 5000 during testing and the no-reordering-overpunctuation heuristic.”
Encoding Source Language with Convolutional Neural Network for Machine Translation (2015, methodology): “For the dependency-to-string decoder, we set rule-threshold and stack-threshold to 10−3, rule-limit to 100, stack-limit to 200.”
On the Impact of Various Types of Noise on Neural Machine Translation (2018, background): “…phrase length, and count bin features (Blunsom and Osborne, 2008; Chiang et al., 2009), a maximum phrase-length of 5, compact phrase table (Junczys-Dowmunt, 2012) minimum Bayes risk decoding (Kumar and Byrne, 2004), cube pruning (Huang and Chiang, 2007), with a stack-size of 1000 during tuning.”
The Edinburgh/JHU Phrase-based Machine Translation Systems for WMT 2015 (2015, methodology): “In decoding we applied cube pruning (Huang and Chiang, 2007) with a stack size of 5000 (reduced to 1000 for tuning), Minimum Bayes Risk decoding (Kumar and Byrne, 2004), a maximum phrase length of 5, a distortion limit of 6, 100-best translation options and the no-reordering-over-punctuation…”
Integrating an Unsupervised Transliteration Model into Statistical Machine Translation (2014, neutral): “See Table 1 Baseline Settings: We trained a Moses system replicating the settings used in competition-grade systems (Durrani et al., 2013b; Birch et al., 2013): a maximum sentence length of 80, GDFA sym-metrization of GIZA++ alignments (Och and Ney, 2003), an interpolated Kneser-Ney smoothed 5-gram ”
Models and Inference for Prefix-Constrained Machine Translation (2016, methodology): “2We choose cube pruning (Huang and Chiang, 2007) as the beam-filling strategy.”
Phrasal: A Toolkit for New Directions in Statistical Machine Translation (2014, background): “The default is the phrase-based variant of cube pruning (Huang and Chiang, 2007).”
The QT21/HimL Combined Machine Translation System (2016, methodology): “In decoding, we use MBR (Kumar and Byrne, 2004), cube-pruning (Huang and Chiang, 2007) with a pop-limit of 5000, and the Moses ”monotone at punctuation” switch (to prevent reordering across punctuation) (Koehn and Haddow, 2009).”
Edinburgh’s Statistical Machine Translation Systems for WMT16 (2016, methodology): “In decoding we applied cube pruning (Huang and Chiang, 2007) with a stack size of 5000 (reduced to 1000 for tuning), Minimum Bayes Risk decoding (Kumar and Byrne, 2004), a maximum phrase length of 5, a distortion limit of 6, 100-best translation options and the no-reordering-over-punctuation…”
Edinburgh’s Submission to all Tracks of the WMT 2009 Shared Task with Reordering and Speed Improvements to Moses (2009, methodology): “Note that this early discarding is related to ideas behind cube pruning (Huang and Chiang, 2007), which generates the top n most promising hypotheses, but in our method the decision not to generate hypotheses is guided by the quality of hypotheses on the result stack.”
The JHU Machine Translation Systems for WMT 2016 (2016, background): “, 2009), a distortion limit of 6, maximum phrase-length of 5, 100-best translation options, compact phrase table (JunczysDowmunt, 2012) minimum Bayes risk decoding (Kumar and Byrne, 2004), cube pruning (Huang and Chiang, 2007), with a stack-size of 1000 during tuning and 5000 during test and the nor”
cdec: A Decoder, Alignment, and Learning Framework for Finite- State and Context-Free Translation Models (2010, background): “, Markov features, but not generally practical for translation), cube pruning, and cube growing (Huang and Chiang, 2007).”
Forest Reranking: Discriminative Parsing with Non-Local Features (2008, methodology): “For non-local features, we adapt cube pruning from forest rescoring (Chiang, 2007; Huang and Chiang, 2007), since the situation here is analogous to machine translation decoding with integrated language models: we can view the scores of unit non-local features as the language model cost, computed…”
Human Effort and Machine Learnability in Computer Aided Translation (2014, methodology): “The default Phrasal search algorithm is cube pruning (Huang and Chiang, 2007).”