Large and Diverse Language Models for Statistical Machine Translation

By Holger Schwenk, Philipp Koehn. 2008

48 citations recorded by the source corpus.

Subfields: Large Language Models, Natural Language Processing

How researchers see this paper

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

In Large Language Models

1. Nadir Durrani
2. Philipp Koehn
3. Eleftherios Avramidis
4. Alexander M. Fraser
5. Helmut Schmid
6. Barry Haddow
7. Hieu D. Hoang
8. Phil Blunsom

In Natural Language Processing

1. Philipp Koehn
2. Nadir Durrani
3. Yonatan Belinkov
4. Barry Haddow
5. Kenneth Heafield
6. Alexander M. Fraser
7. Hassan Sajjad
8. Rico Sennrich

Citations in the corpus

• Edinburgh’s Phrase-based Machine Translation Systems for WMT-14 (2014, methodology): “We trained domain-specific language models separately and then linearly interpolated them using SRILM with weights optimized on the tuning set (Schwenk and Koehn, 2008).”
• Compositional Morphology for Word Representations and Language Modelling (2014, methodology): “Rescoring lattices with a CSLM proved to be beneficial for ASR (Schwenk, 2004) and was subsequently applied to translation (Schwenk et al., 2006; Schwenk & Koehn, 2008), reaching training sizes of up to 500m words (Schwenk et al., 2012).”
• DFKI’s system for WMT16 IT-domain task, including analysis of systematic errors (2016, methodology): “All language models were interpolated on the tuning set (Schwenk and Koehn, 2008).”
• Domain Control for Neural Machine Translation (2016, methodology): “The approaches vary from in-domain data selection based methods (Hildebrand et al., 2005) (Moore and Lewis, 2010) (Sethy et al., 2006) to in-domain models mixture-based methods (Foster and Kuhn, 2007) (Koehn and Schroeder, 2007) (Schwenk and Koehn, 2008).”
• Can Markov Models Over Minimal Translation Units Help Phrase-Based SMT? (2013, methodology): “Parallel We follow the approach of Schwenk and Koehn (2008) and trained domain-specific language models separately and then linearly interpolated them using SRILM with weights optimized on the held-out dev-set.”
• Neural network language models for off-line handwriting recognition (2014, background): “Regardless of the strategy chosen, the combination of both models is a key requirement for improving results [17].”
• QCRI-MES Submission at WMT13: Using Transliteration Mining to Improve Statistical Machine Translation (2013, methodology): “We follow the approach of Schwenk and Koehn (2008) by training domain-specific language models separately and then linearly inter-polate them using SRILM with weights optimized on the held-out development set.”