Search Descriptions


Neural machine Translation

Statistical Machine Translation

Search Publications





Incremental Updating

New training data may arrive continuously, thus allowing incremental updating of models.

Incremental Updating is the main subject of 28 publications. 4 are discussed here.


When a user interacts with machine translation, a desirable feature for the machine translation engine is to incrementally learn from each user translation. Henríquez Q. et al. (2011) incrementally update the translation model and language model with blocks of new translations. Word alignment is carried out with TER, extracting only reliable new translation units. A language model trained on the new data is interpolated with the baseline. Potet et al. (2011) compare incrementally updated translation models against learning a dedicated post-editing model and using the new incoming data for tuning. A related approach is the incremental updating of the translation model from a large data stream (Levenberg et al., 2010).



Related Topics

New Publications

  • Ortiz-Martínez (2016)
  • Mirkin and Cancedda (2013)
  • Denkowski et al. (2014)
  • Dara et al. (2014)
  • Green et al. (2013)
  • Denkowski et al. (2014)
  • Farajian et al. (2014)
  • Mathur et al. (2014)
  • Zhechev (2014)
  • Ortiz-Martínez et al. (2010)
  • Hardt and Elming (2010)
  • Ortiz-Martínez et al. (2011)
  • Gao et al. (2011)
  • Blain et al. (2011)
  • Blain et al. (2012)
  • González-Rubio et al. (2012)
  • Banerjee et al. (2012)
  • Bertoldi et al. (2013)
  • Rocha and Sanchez (2013)
  • Simard and Foster (2013)
  • Wäschle et al. (2013)
  • Cettolo et al. (2013)
  • Mathur and Cettolo (2014)
  • Alabau et al. (2014)
  • Mathur et al. (2013)