The News in Hausa.

Another link from frequent commenter Paul: BBC News in Hausa. As he says, it’s a useful resource because “you know roughly what the news is in English so it should give you good clues.” Another example of the riches provided by the internet; when I bought my Hausa dictionary, I never dreamed it would become so easy to hear the language used! And it was worth investigating the link just for the short clip Boko Haram ta dauki alhakin harin Baga (my dictionary tells me it means “Boko Haram takes responsibility for the Baga raid”), which teaches me how to say Boko Haram (see this LH post) in Hausa: the first word sounds to me like two equally long and high-pitched syllables, the second word is unstressed and lower-pitched.

Comments

  1. Yes! So many languages to listen to. I look forward to all-audio match-the-obscure-language quizzes, and phonology textbooks where all the examples are accompanied by audio.
    I love the chance to hear languages which sound to me like nothing else, like Chatino (even if it’s a ridiculous Jesus movie).

  2. Dictionary use is so old fashioned!

    Google translate tells that this phrase means ” Boko Haram has claimed responsibility for the attack Baga”.

    These days, languages are classified by ability of Google Translate to provide accurate and readable translations from these languages.

    Hausa translations are pretty good, actually, so the language may have high potential.

    Google-translated Japanese in comparison is unreadable to the point of being useless.

  3. Athel Cornish-Bowden says:

    As a more stringent test I tried Google Translate with a longer passage in Hausa (from the same site), about King Abdullah’s death: it was 100% intelligible and reasonably close to natural.

  4. Interesting, I wouldn’t have guessed it would do so well with Hausa. I wonder why?

  5. High-quality bilingual corpora, is the general answer, and zero steps between Hausa and English.

  6. Language structure also appears to be a factor.

    Google translate is very bad with Japanese, but reasonably good with Chinese.

    This obviously has to do with Japanese being agglutinative language, while Chinese is analytic.

    It’s makes machine translation much easier if a word has only one form instead of dozens or even hundreds.

  7. High-quality bilingual corpora, is the general answer

    Yeah, but with Hausa and English?

  8. Trond Engen says:

    Y: all-audio match-the-obscure-language quizzes

    Every Sunday at the Omniglot Blog.

  9. Google translate is also surprising good for Afrikaans. I don’t know whether this is because Google has got its hands on a large trove of bilingual documents or whether it’s due to the relatively simple structure of the language. I suspect it’s more the latter. GT is often very weak with French and one would have thought that the French/English corpora would be the largest by a country mile (UN/EU/Canada/IMF etc.).

  10. -Yeah, but with Hausa and English?

    BBC Hausa service archives, I suspect.

  11. Ah, of course.

  12. Stefan Holm says:

    Google translate is also surprising good for Afrikaans.

    I would be surprised if it wasn’t. English and Afrikaans are the two inflectionally most simplified West Germanic languages, i.e. Afrikaans of all languages is in a sense the closest relative to English there is. The fact that both are spoken in the same country doesn’t exactly complicate the issue.

  13. Athel Cornish-Bowden says:

    I wouldn’t have guessed it would do so well with Hausa. I wonder why?

    More than a year has gone by, and I continue to be very interested (“obsessed” would put it too strongly) in this question. Two ideas that came up recently:

    1. When I mentioned this at work someone said that probably the CIA have lots of people already who can understand French, Spanish etc., so they don’t care if Google Translate does a very poor job with them. On the other hand they need to understand the terrorists in Nigeria, so they put resources into efficient machine translation of Hausa.

    2. The World’s Chief Languages says that Hausa is a strict SVO language (even stricter than English: for example, it doesn’t change the word order in questions). That presumably makes it easy to identify the subject, verb and object, and I wonder if that makes machine translation easier. As we know, trying to translate French or German with a dictionary without knowing the languages produces terrible results. As you have a Hausa dictionary, Hat, maybe you can tell if one can make sense of a Hausa text having only a dictionary as aid?

  14. David Eddyshaw says:

    The main thing is probably that there’s a great deal of written material in Hausa for the Google algorithms to chew on; probably more than any other indigenous African language south of the Sahara. Given the way that Google translate operates, that’s probably of greater significance than any intrinsic feature of the language itself.

    However, for what it’s worth:

    Hausa verbs have plenty of complications of their own to do with numerous derivational-bordering-on-flexional processes which function rather like the binyanim of Hebrew verbs. The relationship between perfective and imperfective stems is not all that predictable either, but of course these sorts of lexical matters are dead easy for a computer, as are the notoriously complex (for humans) Hausa noun plurals. And person, mood and tense marking is very straightforward, being always part of a (separately written) preverbal element, identical for all verbs. You never have to go deducing the subject from context, as you do all the time as in Japanese: exactly the sort of thing computer algorithms struggle wiith.

    Question words in fact usually come first, but this goes with obligatory changes in the verb phrase which basically amount to clefting, so in that sense the word order is indeed more strictly SVO than in English.

    It’s doubtless no more sensible with Hausa than with English to try to account for the huge L2 use of the language on the grounds of “simplicity” rather than historical accident, but it perhaps is possible to maintain that both languages are comparatively easy to speak badly yet comprehensibly and usefully. That could be relevant, maybe.

  15. David Eddyshaw says:

    It occurs to me too that the absence of tone marking in ordinary written Hausa causes much less ambiguity than in Yoruba (say), which probably makes Google Translate’s processes easier. However, I’ve no idea how good or bad GT’s rendering of Yoruba is. (Though when I experimented just now, it told me that “oyinbo” means “cake” …)

  16. As you have a Hausa dictionary, Hat, maybe you can tell if one can make sense of a Hausa text having only a dictionary as aid?

    Nope, at least I can’t — too much morphological complexity.

  17. Athel Cornish-Bowden says:

    I said The World’s Chief Languages. I meant The World’s Major Languages (Bernard Comrie).

  18. numerous derivational-bordering-on-flexional processes which function rather like the binyanim of Hebrew verbs

    No coincidence, that, but a common inheritance from Proto-Afroasiatic.

Speak Your Mind

*