Project Masakhane for machine translation of African languages

[carlos]

I read an article today about some AI technicians working on a project titled Masakhane which has the aim of developing machine translation for all African languages. That would allow many millions of people who speak African languages with very limited literature to access any information that is available in English or French. It would also allow better communication among people from different tribes.

 

One of the obstacles the researchers came across is the lack of written materials in many of those languages in order to feed their databases. So what are they doing? They are using publications from JW.org! They say those are available in a big number of African languages, even some which have very few speakers, and the quality of the text is very good.

 

I read the article in a Spanish newspaper but I found this article in English about the same subject:

https://venturebeat.com/2019/11/27/the-masakhane-project-wants-machine-translation-and-ai-to-transform-africa/amp/

 

[Sheep]

I'm just curious. Where did you hear about them using our publications to translate? That didn't seem to be in the article (unless I completely missed it).

[pnutts]

For example, shortly after launch, Masakhane looked to the JW300 data set of 380 languages from Jehovah’s Witness texts, an insight the group gained following attendance at ACL.

[carlos]

23 hours ago, Sheep said:

I'm just curious. Where did you hear about them using our publications to translate? That didn't seem to be in the article (unless I completely missed it).

Exactly. In the article I read in Spanish the info was a bit more lengthy. But at least that article mentions that they used publications from Jehovah's Witnesses.

[Sheep]

On 12/4/2019 at 6:18 PM, carlos said:

Exactly. In the article I read in Spanish the info was a bit more lengthy. But at least that article mentions that they used publications from Jehovah's Witnesses.

Ah! I did miss it. The article you cited does mention our texts. (Now, how come I didn't see that?)

 

Quote

For example, shortly after launch, Masakhane looked to the JW300 data set of 380 languages from Jehovah’s Witness texts, an insight the group gained following attendance at ACL.

[Dages]

A lot of work in Translation field is now about post edition of machine translation.

You have a computer, an IA, work, prepare a target translation from a source text, then the translator will edit this neuralnetwork production.

Work is quicker.