Search Descriptions

General

Neural machine Translation

Statistical Machine Translation

Search Publications


author

title

other

year

Syntactic Re-Ranking

After a decoder generates the most promising translations, these can be checked for syntactic adherence in a separate stage.

Syntactic Reranking is the main subject of 9 publications. 5 are discussed here.

Publications

Linguistic features may be added to an n-best of candidate translation to be exploited by a re-ranking approach. This was explored by Och et al. (2004) and by Koehn and Knight (2003) for noun phrases. Sequence models trained on part-of-speech tags (Maynard et al., 2007) or CCG supertags (Hassan et al., 2007) may be also used in re-ranking. To check if the dependency structure is preserved during translation, the number of preserved dependency links or paths may be a useful feature (Nikoulina and Dymetman, 2008).

Benchmarks

Discussion

Related Topics

New Publications

  • Barone (2015)
  • Sudoh et al. (2011)
  • Wu and Fung (2009)
  • Rosa et al. (2012)

Actions

Download

Contribute