Boosting Neural Machine Translation from Finnish to Northern Sámi with Rule-Based Backtranslation

Tutkimustuotos: Artikkeli kirjassa/raportissa/konferenssijulkaisussaKonferenssiartikkeliTieteellinenvertaisarvioitu

Abstrakti

We consider a low-resource translation task from Finnish into Northern Sámi. Collecting all available parallel data between the languages, we obtain around 30,000 sentence pairs. However, there exists a significantly larger monolingual Northern Sámi corpus, as well as a rule-based machine translation (RBMT) system between the languages. To make the best use of the monolingual data in a neural machine translation (NMT) system, we use the backtranslation approach to create synthetic parallel data from it using both NMT and RBMT systems. Evaluating the results on an in-domain test set and a small out-of-domain set, we find that the RBMT backtranslation outperforms NMT backtranslation clearly for the out-of-domain test set, but also slightly for the in-domain data, for which the NMT backtranslation model provided clearly better BLEU scores than the RBMT. In addition, combining both backtranslated data sets improves the RBMT approach only for the in-domain test set. This suggests that the RBMT system provides general-domain knowledge that cannot be found from the relative small parallel training data.
Alkuperäiskielienglanti
OtsikkoProceedings of the 23rd Nordic Conference on Computational Linguistics (NoDaLiDa)
ToimittajatSimon Dobnik, Lilja Øvrelid
Sivumäärä6
JulkaisupaikkaLinköping
KustantajaLinköping University Electronic Press
Julkaisupäivätoukok. 2021
Sivut351-356
ISBN (elektroninen)978-91-7929-614-8
TilaJulkaistu - toukok. 2021
OKM-julkaisutyyppiA4 Artikkeli konferenssijulkaisuussa
TapahtumaNordic Conference on Computational Linguistics - [Online event], Reykjavik, Islanti
Kesto: 31 toukok. 20212 kesäk. 2021
Konferenssinumero: 23
https://nodalida2021.github.io/index.html

Julkaisusarja

NimiLinköping Electronic Conference Proceedings
KustantajaLinköping University Electronic Press
Numero78
ISSN (painettu)1650-3686
ISSN (elektroninen)1650-3740
NimiNEALT Proceedings Series
KustantajaUniversity of Tartu
Numero45
ISSN (painettu)1736-8197
ISSN (elektroninen)1736-6305

Tieteenalat

  • 6121 Kielitieteet
  • 113 Tietojenkäsittely- ja informaatiotieteet

Siteeraa tätä