techiaith/moses-smt repository overview

⁠Moses-SMT

(please scroll down for English)

Dyma fodd hwylus iawn i chi osod a defnyddio cyfieithu peirianyddol Moses-SMT yn lleol ar eich cyfrifiadur gyda modelau wedi eu hyfforddi eisoes neu i hyfforddi modelau eich hunain.

⁠Rhedeg gyda modelau wedi eu hyfforddi eisoes

I rhedeg Moses-SMT gyda model cyfieithu sydd eisoes wedi ei hyfforddi, dim 2 orchymyn sydd ei angen.

⁠Gorchymyn 1 : Gosod Moses-SMT

 $ docker pull techiaith/moses-smt

Bydd hyn yn llwytho ac yn gosod isadeiledd cyfieithu peirianyddol o fewn eich system Docker.

⁠Gorchymun 2 : Cychwyn Peiriant Cyfieithu o’ch Ddewis

Mae’r Uned Technolegau Iaith (techiaith) wedi hyfforddi modelau cyfieithu yn barod ar eich gyfer ar sail data rydym wedi’i gasglu o ffynonellau agored a chyhoeddus, megis Cofnod y Cynulliad a’r Ddeddfwriaeth ar-lein.

Mae gan y peiriannau enwau a chyfeiriadau cyfieithu penodol:

• CofnodYCynulliad : ar sail Cofnod y Trafodion Cynulliad Cenedlaethol Cymru
• Deddfwriaeth : ar sail testunau deddfwriaeth DU a Chymru
• Meddalwedd : ar sail cyfieithiadau amrywiaeth o feddalwedd cod agored i'r Gymraeg

Dyma’r ail orchymyn, gan ddewis peiriant ‘CofnodYCynulliad’ a’i osod i gyfieithu o’r Saesneg i’r Gymraeg :

 $ docker run --name moses-smt-cofnodycynulliad-en-cy -p 8080:8080 -p 8008:8008 techiaith/moses-smt start -e CofnodYCynulliad -s en -t cy

Bydd y system yn llwytho ffeil i lawr (tua 3Gb mewn maint yn achos peiriant CofnodYCynulliad) cyn iddo gadarnhau ei fod yn barod i dderbyn ceisiadau i’w cyfieithu.

Os agorwch chi eich porwr a mynd at http://localhost:8008⁠, dylai ffurflen syml ymddangos er mwyn i chi wirio a yw’r peiriant yn gweithio ai peidio.

⁠Moses-SMT

This is a very easy method by which you can install and use Moses-SMT machine translation engines locally with either pre-trained models or to train your own models.

⁠Run with pre-trained translation models

To use any pre-trained model, there are only two commands involved:

⁠Command 1 : Installing Moses-SMT

 $ docker pull techiaith/moses-smt

This will download and install a Moses machine translation system within your Docker environment.

⁠Command 2 : Start a Machine Translation Engine of your Choice

The Language Technologies Unit (techiaith) have trained translation models from bilingual data collected from open and public sources, such as the Proceedings of the Welsh National Assembly, the UK and Welsh Legislature website, as well as localisations of open source software. Each in turn provide domain specific machine translation capabilities. Each one is identified according to its Welsh name. Thus:

• CofnodYCynulliad : as trained from the Record of Proceedings of the Welsh National Assembly
• Deddfwriaeth : as trained from UK and Welsh legislature
• Meddalwedd : as trained from localisations of various open source software projects.

These names can be used in the second Docker command that will start (and fetch is necessary from the Welsh National Language Technologies Portal) a translation engine for a desired source and target language pairing:

 $ docker run --name moses-smt-cofnodycynulliad-en-cy -p 8080:8080 -p 8008:8008 techiaith/moses-smt start -e CofnodYCynulliad -s en -t cy

In the case of CofnodYCynulliad, the engine may be a very large download - about 3Gb.

Open your browser and browse to http://localhost:8008⁠, where you should see a simple demo form that will help you check if the engine is working or not.