110 New Languages at Google Translate.

Google announces:

Google Translate breaks down language barriers to help people connect and better understand the world around them. We’re always applying the latest technologies so more people can access this tool: In 2022, we added 24 new languages using Zero-Shot Machine Translation, where a machine learning model learns to translate into another language without ever seeing an example. And we announced the 1,000 Languages Initiative, a commitment to build AI models that will support the 1,000 most spoken languages around the world. Now, we’re using AI to expand the variety of languages we support. Thanks to our PaLM 2 large language model, we’re rolling out 110 new languages to Google Translate, our largest expansion ever.

From Cantonese to Qʼeqchiʼ, these new languages represent more than 614 million speakers, opening up translations for around 8% of the world’s population. Some are major world languages with over 100 million speakers. Others are spoken by small communities of Indigenous people, and a few have almost no native speakers but active revitalization efforts. About a quarter of the new languages come from Africa, representing our largest expansion of African languages to date, including Fon, Kikongo, Luo, Ga, Swati, Venda and Wolof.

Fon in the news again! But what stood out to me was the inclusion of NKo in the further list of newly supported languages; I had thought it was an alphabet (as discussed here in 2011), but apparently it’s also a language:

NKo (ߒߞߏ) is a standardized unified koiné form of several Manding languages written in the NKo alphabet. It is used in Guinea, Guinea-Bissau, Mali, Ivory Coast, Burkina Faso and some other West African countries, primarily, but not exclusively, in written form, whereas in speech the different varieties of Manding are used: Maninka, Bambara, Dyula and others.

It is a literary register with a prescriptive grammar known as kángbɛ (“clear language”) codified by Solomana Kante, with the màninkamóri variety, spoken in Kante’s native Kankan region, serving as the mediating compromise dialect.

Thanks, Martin!


  1. Bathrobe says

    Well, this is all well and good. The proof of the pudding is in the eating. It would be nice, though, if Google Translate could fix its problems with existing languages. And some of them are pretty bad.

    For Mongolian, the translation of grammatical terminology is completely abysmal (a total balls-up, if I may say so). Now I understand that grammatical terminology is pretty esoteric in any language. But Google Translate takes the snafu to a whole new level. Surely somewhere in the screeds of raw data that Google Translate works on the algorithm must have picked up some terminological equivalents. But no, the translation that Google Translates gives aren’t just out of the ball park, they are often totally and completely and utterly wrong. (I’m temporarily in China so I can’t access Google Translate to provide you with examples.)

    To make things worse, the Google Translate team is completely impervious to feedback. (But this includes Google as a whole. I informed them two months ago that the information on applying for a Chinese visa in Mongolia is outdated but Google Search still returns the same outdated and extremely misleading result. But I digress.)

    The article suggests that many of these languages are primarily spoken idioms. Well, that might help, since many of the languages given probably don’t even have a term for, say, “definite article”. But a machine learning model “that learns to translate into another language without ever seeing an example” looks like a recipe for an even bigger disaster to me. Even in the field of ordinary language Google Translate can be left totally at sea. Try translating “I want to go”, “I want you to go”, and “I want him to go” into a language (like Mongolian) that doesn’t use structures analagous to English. To be frank, some of the translations are worse than useless.

    So I am not optimistic about Google’s new foray.

  2. David Eddyshaw says

    The Twi one is crap. I think a lot of these are basically window-dressing to give the impression that Google is hip and happening rather than an evil incarnation of surveillance capitalism.

    The “AI” boosterism fits the general PR-over-fact strategy too. It fits another pattern too, that we’ve seen two examples of on LH lately: high-pressure salesmanship of the benefits of “AI” to African literacy programs. Much preferable to, like, training local teachers and educators, employing them. paying them, listening to what actual ordinary African want … and what their actual needs are. American tech billionaires will solve your problems …

    Google has no incentive to respond to complaints about its search algorithms; quite the contrary:


  3. Bathrobe says

    Couldn’t open the link to doctorow, but that could be due to the Great Chinese Firewall. I tried searching on Baidu but no luck.

    Baidu did, however, give a translation for “the specific process by which Google enshittified its search”:


    Roughly translated: The specific process of Google consecrating its search.

  4. David Eddyshaw says

    Well, it’s nearly right …

  5. It would make for a rather boldly innovative (to use two common buzzwords…), brand-new kind of comparative linguistics: how poorly does Google Translate handle different language pairs, and more interestingly, what factors correlate with (possibly cause) observed differences in quality? I would imagine structural and semantic differences must play a massive part, as does the size of the available written corpus for each language, and how uniform/standardized each such written language corpus is. On the last point I was floored when I saw that they seem to treat Berber as a single written language: I suspect Google Translate into and from Berber will be atrocious if not mostly useless (Can any hatter with knowledge of Berber -Lameen?- confirm this?).

    David Eddyshaw: Not only do I agree with your assessment that this is (mostly) simply window-dressing, I strongly suspect that their including some languages which are extinct but with “active revitalization efforts” is for purposes of P.R., inasmuch as it would be difficult to criticize the quality of Google Translate in instances where the language lacks native speakers -thus, claims on the part of Twi speakers (that the system does not work) will be “balanced” by claims that it does work adequately -made by speakers of an extinct language!

  6. DE, who and when ever listened to what people want?

    What is good about developers of translation AI and teachers is that some of them at least do what they want. Many (teachers at least) are even Africans. But… as you noted, AI is cheaper:)

  7. So, since now I won’t be coming across langauges not in the list?

    Tamazight (Amazigh) is a Berber language spoken across North Africa. Although there are many dialects, the written form is generally mutually understandable. It’s written in Latin script and Tifinagh script, both of which Google Translate supports.

    So Tamazight is just Berber (and not for example those Moroccan varieties whose speakers call their langauge so).

    ” opening up translations for around 8% of the world’” – what about Russian speakers who want to read a text in something else than English? Why not 100%?

  8. Touareg is absent though. Maybe it is included in Tamazight?
    It seems to be easier to just add a couple of sufficiently popularised (in printed form) Berber languages. Like Kabyle.

    Adding each variety separately means first taking responsibility for compiling the list of “varieties” (and a lot of effort to compile the corpora for training) adding them all together must pose interesting technical problems (again corpora and only).

    Apart of interesting problems, there is an issue of translation to Tamazight. Will GT invent an intermediary langauge of its own?

  9. I think a lot of these are basically window-dressing to give the impression that Google is hip and happening

    That would not surprise me even a little.

  10. “I suspect Google Translate into and from Berber will be atrocious if not mostly useless ” – Arabic is mostly useless (butI haven’t tried it for ages).

    I tried the main page of new “Amazigh Wikipedia”, https://zgh.wikipedia.org/wiki/ⵜⴰⵙⵏⴰ_ⵏ_ⵓⵙⵏⵓⴱⴳ
    It must be easier than a transctipt of a conversation, I suppose?

    It detected “English” and translated it from English to English. So I manually switched it to “Tamazight (Tifinagh)”. Same with “Tachelhit” (switched to “Tamazight”). It confused Kabyle with a distantly related Afroasiatic language “Arabic”.

  11. Results: Tamazight, Kabyle, Shilha.

    Oh. My links still are trying to auto-detect the language….Well, I won’t fix them:)))))

  12. Or I will: Tamazight, Kabyle, Shilha (fixed).

  13. “Brrk on Wikipedia. The free adrfi language that everyone can write in !” (Shilha)

    To be fair, all three are readable (with many errors).

  14. Tamazigh:
    “The Cauchy–Schwarz equation says that for all two waves u is v of a planet with a natural product here will be darnɣ tili
    |⟨u,v⟩|² ≤ ⟨u,u⟩⋅⟨v,v⟩
    but ⟨⋅,⋅⟩ it is a national product. For example, a social product includes a physical product of colors and materials.”

  15. “readable (with many errors). ” – Yes. But WP is easy. Editors of Russian WP sometimes introduce Anglicisms on purpose (because en.WP is the source of all knowlege) – Kabyle editors must be doing soemthing similar with French and Arabic.
    It tried the song that I just mentioned elsewhere… It’s an easy text as well. A half of lines are translated correctly (here I must rely on other translations e.g. by Brugnatelli) another half is screwed up. And how exactly depends on line-breaks and capitalisation (and of course transcription: B.’s transcription in the article above slightly differs from the spelling used in WP).

  16. Cf. B.’s transcription and translation (with actual line breaks, not “/”):
    Adfel yessud tibbura / tugi kecmen-tt yeḥlulen / tajmaɛt tettargu tafsut / aggur d yetran ḥejben
    La neve si accumula contro le soglie / nella pentola è entrata la minestra invernale / la piazza del paese sogna la primavera / la luna e le stelle sono velate
    GT: The snow soaked the windows. / she refused to enter it / congregation waiting for spring / the moon and stars are beautiful

    WP: Adfel yessud tibbura, tuggi kecment yeḥlulen Tajmaɛt tettsargu tafsut aggur d itran ḥejben
    The rain has covered the doors, the sweet ones have refused to enter. The congregation is waiting for spring, the moon and the stars are bright
    Adfel yessud tibbura, tuggi kecment yeḥlulen / Tajmaɛt tettsargu tafsut aggur d itran ḥejben
    The rain closed the doors, the sweet refused to enter / The community is waiting for spring, the moon and the beautiful stars
    Adfel yessud tibbura, / tuggi kecment yeḥlulen / Tajmaɛt tettsargu tafsut / aggur d itran ḥejben
    The snow covered the doors, / refusal to enter the solution / The community is waiting for spring / The moon and stars are beautiful

  17. About N’Ko: the first of the three references in the article about Sulemaana Kantè tells that

    1) when he first saw Ajami Maninka and could not understand it, he “said African languages are worthless because God has not allowed for them to be written”.

    2) when he next saw a book by a Lebanese author telling how African languages have no grammar he took offence and began to work on Manding literature (with the purpose of spreading Islamic knowlege, hello Protestants).

    By the way, kángbɛ (“clear language”) in his books refers to grammar:

    Because mastering a language in writing is very hard, experience has shown that every language has its rules. Grasping a language’s rules facilitates knowing its writing. As such, people created explanatory books. These books clarify the language properly, remove blemishes from it, and make knowing the language much easier! The name of the book of established rules of a language is ‘kángbɛ’.

    From the third link.
    (though Delafosse calls so the koine, and certian Jula speakers call so their language. I wonder if this name is (a) endonym [among its speakers] only or an exonym too (b) contrasts it to other forms of Manding or to other languages. If it is an endonym then it can be similar to Speaking Simply (that is, not in Lithuanian) which used as a dialect name for what Locals speak)

  18. The third contains this convenient table (I don’t know whether it is accurate):

    Local Name     Etymology       French Name English Name Alternative
    màndinkakán “Language of the    mandingue, Mandinka,
                 people of Manden”  malinké    Mandingo
    màninkakán  “Language of the    malinké    Maninka
                 people of Manden”
    bàmanankán  “Language of those  bambara    Bamanan      Bamana
                 that refuse (Islam)”
    jùlakán     “Trader’s language” dioula     Jula         Dyula, Diula,

  19. 3d ref again:

    Though it is beyond the scope of this chapter, it is important to note that for Kantè, N’ko is not his baptismal name; it is rather an archaic name for the language that was used as far back as the founder of the Mali Empire, Sunjata Keïta in the 13th century

    So where “N’ko” comes from?


    The language which the Mandings speak is N’ko.
    Màndén’ nù lá Ń’ko’ yé kànbolón’ kùnbabá’ 4 nè dí. Ò lù fɛ́ lɛ́ nìn: (bànbàran, mànènka, màndènko, à ní jùla)
    The Mandings’ N’ko is 4 principal dialects. Take a look at them:
    (Bamanan, Maninka, Mandinka, and Jula).

  20. David Eddyshaw says

    when he next saw a book by a Lebanese author telling how African languages have no grammar he took offence and began to work on Manding literature

    (I suspect that “literature” is a mistake, for “literacy.”)
    The “Lebanese author” was, one imagines, simply an ignorant bigot.

    This is rather reminiscent of the business about Blackfoot we discussed here a little while back. Knowledge doesn’t count if it’s not from an approved source. Kante was labouring under a number of serious misconceptions about his own language, and African languages in general.

    “African languages are worthless because God has not allowed for them to be written”

    This is presumably what it was all about for him. Hence the equation of “writing system” with (proper) “language”, resulting in using the same neologism (“I talk”) for the writing system and his artificially created interlanguage. (Not that I have anything against artificially created interlanguages, so long as they know their place and don’t claim to be Primeval High X-ian.)

    There is a similar thing going on with the excellent (and genuinely charming in his enthusiasm and forebearance with our doubting) Almeida Samo


    Not a few Europeans and Americans labour under similar misconceptions about the relationships between speech and writing.

  21. when he next saw a book by a Lebanese author telling how African languages have no grammar

    I wonder, was it this chapter, “Negro Languages” (لغات الزنوج), in Kamel Mrowa (كامل مروة)’s We in Africa (نحن في أفريقيا)?

  22. @DE, “literature” is my word, I’m just retelling what I read. And I should have written “written literature”.

    Well, honestly, written literary tradition entails literacy and literacy entails literature (one can imagine one wihtout the other, especially now when we have AI to write and read for us). But:

    “Critically, Kantè also used his unique script to write over 100 books on a vast range of topics spanning across linguistics, history, traditional medicine and Islam (including a translation of the Quran), which continue to be typeset and sold alongside the works of current N’ko intellectuals today.”

    (ref3 in the WP article)

    I think I just wanted to emphasise this and also what I read in ref 1 (that the Quran in Manding helps understand Islam).

  23. David Eddyshaw says

    The WP article is brief:


    It mentions that he translated Mein Kampf into Arabic for an antisemitic newspaper, which suggests that he might have been expected to have a somewhat unenlightened outlook on Africa.

    There is (or was) a big Lebanese diaspora in West Africa, largely of merchants and traders, retaining strong ties to Lebanon; relationships with the locals have not always been happy. (I knew one – a diamond merchant – who was actually murdered in Togo.)

    (The Bad Guy/Tempter, Yusef, in Greene’s The Heart of the Matter, set in Sierra Leone, is described as “Syrian”, but I imagine he would probably have been Lebanese if he’d been real.)

  24. MMcM, wow! Thanks for identifying it.

    Yes, the author transcribes him as “Kamal Maruwa [Kamal Marwa]” (I think, a transcription from N’ko [and Arabic])

    The “Lebanese author” was, one imagines, simply an ignorant bigot.

    DE, when you’re accustomed enough to various epithets Arabs give to their vernacular you can’t say if this is bigotry or just the very same position without sparing smaller peoples and less developed countries (as modern Arabs would do when insisting that “slang” which they speak themselves “has no grammar”). But thanks to MMcM we now can just read it.

  25. David Eddyshaw says

    Just because people are ignorant bigots about actual spoken Arabic doesn’t mean that their attitude to (other) African languages isn’t ignorant bigotry. Ignorant bigotry is pretty adaptable.

    Still (as you imply), Kante may very well have imbibed from such people the notion that nothing actually has, like, grammar except Fuṣḥā. That would easily feed into his idea that only written languages have grammar. (As I say, this delusion is by no means unknown in Anglophonia. It has only recently bitten the dust in Welsh, too: it was routine to describe actual spoken Welsh as “corrupt”, as opposed to Cymraeg Llenyddol, even a few decades ago.)

  26. There is (or was) a big Lebanese diaspora in West Africa,
    Still exists. Pertaining anecdote: when I worked in Lebanon 15 years ago, we did a bus tour of the North, and one of the highlights was a village whose inhabitants lived in Western Africa and came home only for about two months each year, when they would show off their wealth to each other at banquets. They also built extravagant villas to show off; I remember one of them was built in the form of a Boeing airliner.

  27. “Still (as you imply), Kante may very well have imbibed from such people the notion that nothing actually has, like, grammar except Fuṣḥā. That would easily feed into his idea that only written languages have grammar.”

    DE, this is not how the author understands him: “It was true, Kantè said, Africans didn’t have a writing system, but it was insult to injury to spread the lie that their languages were deprived of grammar.

    And apparently he changed his mind about what God “allowed” when he read Mrowa:))))

    Also in the quotation above he says that the name of the grammar book of a langauge is “kángbɛ” – not the same as “only written languages have grammar” (if this passage is what you extrapolated).

    DE, now Mrowa aside (the guy wasted his time on Hitler’s writing…) and speaking of Arabic and Arabs.

    It is logical that
    (1) different peoples and cultures differ in associated virtues and sins (not to say about qualities which are neither virtuous or sinful)
    (2) when you contact them you’ll be enraged by absence of virtues you believe to to be the norm (and maybe even minimal requirement for basic respectability) and presence of sins which you are not accustomed to so much that you barely notice them and believe to be in human nature.

    And then (3) if I contact other cultures only to reassure myself that the only good people around are educated half-Jewish Muscovites, that would be an Epic Fail. It is not with this purpose in mind I’m so curious about them:) So, (4) one needs some protection frome effects of 2, beforehand.

    Now, logically, again (5) the exact relation between the two (usually three or more) languages Arabs use matches absolutely nothing in my Russian experience.
    It means, this is one of those parameters where peoples differ. Their prejudices (and of course it is normal for Russians too to have some language prejudices) are going to be surprising. And until I’m able to look at the situation (the relation between vernacular and literary) with their eyes – and I still can’t look at language with their eyes – I better abstain from evaluating them.

    Just logic:) Note that all of my Arab freinds LOVE dialects. Seriously. They don’t look at them as some sort of uneducated speech to be replaced by Fus7a. But in certain context they may says something similar.

    Meanwhile “bigot” is a strong word, usually it means more than the sort of prejudices like everyone has. Yes, Marwa, likely is a bigot.
    (not that you are anyhow attacking them! I”m just explaining why I hesitate to evaluate what I do’t really understand)

  28. David Eddyshaw says

    I don’t think it’s a stretch, or any kind of cultural or moral absolutism, to describe Nazi sympathisers as “bigots.” They’re the very reason why we have such words.

  29. DE< I'll repeat then:

    "Mrowa aside (the guy wasted his time on Hitler’s writing…) and speaking of Arabic and Arabs.

    Yes, Marwa, likely is a bigot."

    I’m only speaking of what Arabs normally say about vernacular Arabic in certain contexts (and again, modern Arabs who I know don’t say the same about other languages). Which is much the same as what Mrowa said about African languages [other than Arabic].

  30. @DE, I’m not saying that Mrowa is not a bigot. I’m just noting (I can’t not note) that his words are identical to what you hear from Arabs, including linguists, including people who love dialects, including a Moroccan who loves to insist that “Arabic languages” (plural, and we’re talking about European words here) is better than “Arabic dialects” and who when I mentioned the proposal to print in Moroccan darija and use it in education said that’s ridiculous, it is not even a langauge.

    As a Russian (a langauge where educated speakers are expected to speak the “literarary” Russian! Seen as less diglossic than English with its ubiquotous contrast between “written and spoken [informal, ….]”) you can make sense of this diglossic coordinate system by assuming for example, that both registers are “incomplete” (you won’t spare Fuṣḥā).
    Or else within this system you can decide that African languages are not really languages, but can and should be developed to become langauges. This is what so many Africans and Europeans likewise do.
    Or else within this system you can decide that African lnaguages are not really languages and can/should not be developed: Africans should be taught in European. This is the option chosen by Mrowa (“just can’t”).

    I guess a modern Arab with a degree just won’t discuss African langauges within the Arabic coordinate system, and reserve it to Arabic.

  31. A horrifying overview of horrifying variety of names for Manding on p. 8 (by Vydrin):

  32. Thanks, that explains why I’ve always been so confused about the names of the “Mande” languages and the group as a whole. What a mess!

    I was amused by this blast from the past at the end of the piece:

    To the end of January 1996, I can be reached by e-mail at <vydrine@msu.edu>. My e-mail in Russia is <azic@evrdom.spb.su> but please do not use this for very long messages because the user must pay for each kb of message.

  33. Yes, me too.

    1996 is when I got my first address. I did not have to pay for anything, but still to have an address I needed to be affiliated with (or have freinds in) one of those still few organisations that could offer those. I think should have kept it, it was fun ([name]@[surname].mccme.rssi.ru)

  34. About “Mande”, I’m still trying to figure it.

    French WP says
    – about the region: “Le Mandé, Manden ou encore Manding est une région située en Afrique de l’Ouest…”
    – about “langues mandées”: “Le terme « mandé » fait référence à l’ancien Empire du Mali d’où ces langues sont supposées provenir. Il s’agit d’une construction de savants européens, initiée en 1854 par le missionnaire allemand Sigismund Koelle dans son Polyglotta Africana. Il y mentionne treize langues sous le titre « North-Western High-Sudan » ou « mandenga ». « Mandé » provient d’une segmentation erronée du terme « mandenka », « habitant du pays Manden » (Koelle a analysé -nka comme le suffixe, alors que le suffixe signifiant « habitant » est simplement -ka)[1]” where [1] is this letter from Vydrin.

    But Vydrin says: ‘”Mande” is an artifical creation of European scholars; it was first used in 1867 work by H. Steintal and fixed in its modem “European” meaning by Maurice Delafosse. This form is obtained by the dropping of the final nasal of the word “Manden,” and is not used by speakers of these languages, except for some Bamana or Jula-Konya dialects where the final nasal can be omitted in the citation form but not in combined words.’

    I recently downloaded Koelle (weirdly I only could find it on libgen).
    He does extract -n°ka (“n°” for [ŋ] and it should be dot, not circle) from several of -V / -Vn°ka pairs.
    Perhaps it has to do with citation forms?

  35. David Eddyshaw says

    Denis Creissels in Le Malinké de Kita derives the various forms from *mànden-ka-kan. The *-kan part is uncontroversially “language”; he gives the middle bit as *-(n)ka “habitant de.”

    He goes on to say that Manden (sic) is a fundamentally geographical term, and that “Mali” (as in the empire) is the same name, mangled by neighbouring non-Mande speakers.

    There seems to be no doubt but that the echt form is Manden, with the final n. (This is the only possible syllable-final consonant in Manding: it’s actually a nasal unspecificied for position, very like the Japanese syllable-final n. It can end up as just vowel nasalisation sometimes.)

    I think quite a lot of West African language names are relatively recent inventions. Traditionally, people tended to feel no need to specify more than “our language” or “language of such-and-such a place/people.” There was no need to have any sort of overarching name for what modern linguistics calls a “language”, conceptualised in terms of mutual comprehensibility. “Akan”, for example, was a deliberate modern neologism, created to capture the fact that speakers of Twi, Fante and Akuapem can understand each other well enough.

    Kusaal regularly forms language names from ethnonyms, but where the resulting terms end up corresponding to modern linguistic ideas of a “language”, that’s basically chance. (I think Nabit, for example, is probably actually closer to Toende Kusaal than Agolle Kusaal is.)

    The Yansi apparently used to speak Dyula, but now speak Mooré; no matter – in Kusaal, they have always spoken Yaan.

    And the Nasaal spoken in Burkina Faso is not mutually comprehensible with the Nasaal spoken in Ghana, though both are the language of the Nasaarnam “Europeans.”

  36. “our language” – yes, I wonder if “clear language” is just one of those names.

    for what modern linguistics calls a “language”, conceptualised in terms of mutual comprehensibility.

    I reiterate that “modern linguistics” should not call small language continuums so. Small language continuums are not very different from large ones, and any language continuums are “language continuums”. Speaking of “a language” in this sense of “a small language continuum” as if it they are qualitatively different from other things which are not small language continuums only creates confusion.
    It is an attempt to present a particular approach to language planning as “descriptive” science and as such is…

  37. David Eddyshaw says

    Nahuatl … (though they did say Mexicatlatolli sometimes.)
    And of course, in Inuit you speak Realpersonese. (Or Skraelingese, in some cases.)

    Julakan is literally “Traderspeak.” (Jula is, of course, entirely unrelated to Jola. Just so we’re clear.)

  38. The oldest tradition of Islam among Manding speakers seems to be traceable to the jùlá network that originated first with Muslim Soninke traders that spread out across West Africa during the Ghana empire that preceded that of Màndén (see Wilks 1968, 2000) (See Figure 2).
    During the Mali empire, which reached its apogee in the fourteenth century, the Muslim Jula network of traders became increasingly Manding; that is, older Soninke members adopted the language of Mali and were additionally joined by other Manding-speaking Muslims along their trading routes and outposts (Massing 2000).

    Says the first of two articles referenced in WP, Solomana Kante. I wodner how reliable is this story.

  39. David Eddyshaw says

    Trade and Islam have been intimately linked in West Africa pretty much all along: in fact, trade has done much more to spread Islam than any jihad.

    It’s still the case. Several of our professional drivers from the (Presbyterian) hospital in Bawku found it useful (if not indispensible) to have Muslim personae to slip into on longer trips. (There’s nothing hypocritical about this: a lot of West Africans inhabit several cultures and languages as needed as they go about their lives, often using different personal names depending on the particular cultural or linguistic hat they are wearing that day. So do we all, I suppose, but in Europe we live in much more of a monoculture, so it’s less evident.)

  40. Yes, it makes sense – that’s why I wonder if it is actually supported by sources. It is just the interaction of Christians and Muslims which used to fully consist of mutual conquests. I love the term I once saw in reference to spread of Islam in Siberia: странствующий шейх, “shaykh-errand”.

  41. David Eddyshaw says

    As I’ve mentioned before, the consolidation of Islam in what is now northern Nigeria mostly took place in the colonial period. Although many towns (like the seven traditional Hausa city-states) had been Muslim for centuries, the countryside (where the great majority of people actually lived) was another matter: it was the imposition of the Brit desolation-peace that created favourable conditions for the spread of Islam, which in West Africa has always spread largely through peaceful means, not conquest. (The way to end up on the receiving end of a jihad was generally not to be pagan, but to be The Wrong Kind of Muslim.)

  42. Well, there is a darker parallel to your story about Hausa here. Denham – who won popular diapproval in Europe because he accompanied a slaver raid* – describes a scene where the sultan gives a speech about promotion of Islam and the leader of a party of Arab slavers notes that the sultan is not interested in converting pagans for the obvious reason of slave trade.

    *Or not exactly. Instead of sending them against relatively helpless pagans the sultan tried to emply their company in his warfare with Fula and they were successfully ambushed in a Fula village.

    PS “Offerings poured in, from all the Kerdy nations; and the sultan excused himself to Boo-Khaloom for the delay, on account of the extreme tractability of the people around him, who, he said, were becoming Musselmans without force. Again Musgow was mentioned; adding, that the warlike arm of the Arabs, bearing the sword of the Prophet, might turn their hearts. This hypocrisy, however, Boo-Khaloom inveighed against most loudly to me, declaring that the conversion of the Kerdy people would lose him (the sultan) thousands of slaves, as their constant wars with each other afford them the means of supplying him abundantly.” here

  43. Now after rereading it I’m not sure if it is sarcasm or a complaint. Boo-Khaloom himself is not (materially) interested in promotion of his religion.

    P.S. “Kerdy” must be the word whose etymology we discussed (“Shuwa Arabic kirdi, Kanuri kə́rdì…”) in the other thread. Also WP

  44. David Eddyshaw says

    Yes, there were a number of economic interests in not converting the peasantry to Islam, and not just the slavery thing; also taxation, and military service: pagan Hausa were generally felt to be Just Better as soldiers. Also, no sensible feudal ruler is going to be in a great hurry to alienate his agricultural base by attempting large-scale conversions of the not-particularly-willing, even assuming he has the resources to try.

    So much so, that theological excuses were sought for (and found, naturally) for not doing so: I’ve previously mentioned the cheeky adoption of Maguzawa “Magians, Zoroastrians” for the people-previously-known-as-Arna “pagan Hausa.” Honestly, it’s enough to make you understand where the jihadists were coming from. They did have a point (many points, in fact.)

  45. Bathrobe says

    I reiterate that “modern linguistics” should not call small language continuums so.

    …. which brings us back to the celebrated topic of whether we (or linguists in general) should call non-mutually-intelligible mega-continua like the Sinitic languages “Chinese”.

  46. Bathrobe, now I wonder whether it is a continuum…
    I mean, of course it is matter of degree of continuity. But it would be interesting to know where transitions are smooth.

  47. David Marjanović says

    There is research showing it’s a continuum. Unfortunately it comes out of China, where people have a strong incentive to find that result.

    Given Romance, however, I wouldn’t be surprised at all to find that most of it, perhaps even all of it, is a dialect continuum.

  48. Bathrobe says

    To return to the original topic:

    Now that I’m back in Mongolia I can use Google Translate. Here are a couple of translations it throws up:

    Relative clause: харьцангуй заалт ‘relative (not absolute) (contract) clause’
    Adjectival clause: нэр үг ‘noun’
    Adnominal clause: адноминаль заалт ‘adnominal (contract) clause’
    Modifying clause: өөрчлөх заалт ‘changing (contract) clause’
    Definite article: тодорхой өгүүлэл ‘clear composition/article/piece’ (the ‘clear’ part is actually accurate; it’s the translation of ‘article’ that is weird)
    Subordinate clause: дэд өгүүлбэр ‘sub-sentence’ (this is actually reasonably ok)

    I have also found a Mongolian expression that comes up as ‘definite article’ in English but I can’t remember what it was.

  49. Some 15 years ago people* in South Africa were having fun asking GT to translate variants of “men are men and men must [verb]” (just in Afrikaans) with various verbs.
    GT was translating it as “men are men and women must …. ”
    (tried it recently, the trick does not work now)

    *A assume white, male, but who knows:)

  50. I would have thought “language continuum” is only salient if the dialects at opposite ends of the continuum are not mutually intelligible. A language continuum so “small” that all its dialects are mutually intelligible is just a language, no?

    Of course, intelligibility is itself a continuum…

  51. Well, that’s certainly not the case for Chinese.

  52. I just searched for ‘Zero-shot machine translation’ in Google, and was helpfully supplied with the supposed Latin translation of ‘zero-shot machine’ – “Nulla-iaculat Machina”. See me after class, Google.

  53. I remember now.

    Google translate translated “тодотгол гишүүн өгүүлбэр” (modifying clause, relative clause) as “definite article”. I was gobsmacked.

  54. Speaking of stupid jokes:
    “Когда после слова стоит член…” means in Englsh “when there is an article after a word…” (especially in books about Slavic languages: those about European languages can borrow a local word).

    But modern Russian speaker will read it as “when there is an erect penis after a word….”. What happened:
    (1) for “is” we use here “stands”. “When after word stands член”. You see this “stands” all the time. But “stands dick/dick stands” is also how we say “erection”. (I use “stands” for both continuous and habitual meanings, these are not distinguished)

    (2) since about 70s-80s член “member” became the usual colloquial word for penis (though I personally find it extremely ugly and don’t use it).
    We don’t discuss articles in books about Russian grammar: we don’t have any. In books about English grammar it is artikl’ (and I think for a long time if not since ever because see above about borrowings local words).
    Which means, when a linguistics student (a Slavi[ci]st) encounters the word “member” in the sense “article” for the first time (and most of Russians never encounter it), she’s an adult already. It will be difficult for her to think of any other (unfamiliar) reading.

  55. Bathrobe says

    Google Translate gives “определенный артикль” for “тодотгол гишүүн өгүүлбэр”, so no problem there! (Except the translation itself, of course).

  56. @Graham: GT is notoriously useless for translation into Latin, and doesn’t seem to have improved over time.

  57. Latin literacy amongn LLMs has experinced steady decline since the time of Caesar.

  58. David Eddyshaw says

    Sad but true. They just don’t make LLMs like they used to. Eheu fugaces …

Speak Your Mind