Search Descriptions

Main Topics

Search Publications


author

title

other

year

Phrase-Based Models

Phrase based models, the workhorse of current statistical machine translation, go beyond word based models by mapping sequences of words.

Phrase Based Models and its 10 sub-topics are the main subject of 248 publications.

Publications

The modern statistical phrase-based models are rooted in work by Och and Weber (1998); Och et al. (1999); Och (2002); Och and Ney (2004) on alignment template models. These models defined phrases over word classes that were then instantiated with words.
Translating with the use of phrases in a statistical framework was also proposed by Melamed (1997); Wang and Waibel (1998); Venugopal et al. (2003); Watanabe et al. (2003). Marcu (2001) proposes the use of phrases within word-based model decoding. The use of log-linear models was proposed by Och and Ney (2002).
An influential description is presented by Koehn et al. (2003), which is similar to the model by Zens et al. (2002); Zens and Ney (2004). Tribble et al. (2003) suggest the use overlapping phrases. Lopez and Resnik (2006) shows the contribution of the different components of a phrase-based model.

Benchmarks

Discussion

New Publications

  • Nishino et al. (2016)
  • Cuong and Sima'an (2014)
  • Pal et al. (2011)
  • Junczys-Dowmunt (2012)
  • Wisniewski et al. (2010)
  • Tambouratzis et al. (2011)

Actions

Download

Contribute