From my diary

Last week, on Ash Wednesday, I happened to read some crazy claim by a neopagan that Ash Wednesday derived in some weird way from Woden (!).  Since then I have been working on a post about the origins of Ash Wednesday, and specifically the imposition of ashes.  It’s been a long and weary haul, as I have to work with Dark Ages sacramentaries, but I’m gradually getting there.  Yesterday I discovered the old DACL encyclopedia article on Cendres, which is proving very useful.  It is interesting to see that modern scholarly literature is often unaware of it.  I hope my own article will be done by the end of the week.

March 5 was St Piran’s Day, and I wrote something about that yesterday.  But doing so caused me to retrieve the Latin text of the Vita of St Piran. This was published by Capgrave in Nova Legenda Anglie, at some remote date, and this in turn was nicely reprinted and edited by Horstman in 1901.  Looking at it, I felt that old urge, and fired up Abbyy Finereader 15, and scanned the 9 pages.  Today I finished correcting this to produce a Latin text.  The orthography is dreadful, and a real barrier to the non-specialist.  I’ve corrected some of it in the Word document.

On a whim I pasted it into Google Translate.  I was astonished – but delighted – to find that I got back something very readable indeed:

Google Translate on the Life of St Piran

The translation engine for Latin is clearly going great guns.  There are a few mistakes, but not many at all.  If this is now the standard for medieval Latin, then we all need to get out there and start using it and start producing cleaned-up translations of medieval saints’ lives.  My own hitherto faint urge to translate the Life of St Piran has just received a boost.  On Thursday I might well take a look at doing this.


8 thoughts on “From my diary

  1. “Ash Wednesday derived from Woden”

    That’s got to be some sort of crossed wire where the “Woden” in “Woden’s Day” (Wednesday) has been confused with being a source of Ash Wednesday–right? Or am I now doomed to see “articles” on this newfound knowledge being earnestly shared across the length and breadth of the internet henceforth?

  2. Google Translate used to produce gibberish from Latin texts. It has improved greatly, depending on the source text, but is still too unreliable to use with much confidence. DeepL, I find, is much superior for the languages it has, but unfortunately Latin is not one of them (I’ve used it for Dutch and German and got much better results than from Google). Here is the very faulty Google translation of Tantum ergo:

    So the only mystery
    We bow with reverence
    The old document
    Newer rite:
    Loyalty Supplier
    The lack of senses.

    Father and the Son
    Praise and jubilation
    Health, honor, and virtue
    May it be a blessing
    Proceeding from both
    Praise is not equal.

    I don’t know what the technology is, but Father for “genitori” and Son for “genito” (paraphrases or explanations rather than translations of these terms, which mean the Begetter and the Begotten) probably mean that it is searching existing translations rather than doing grammatical/lexical analysis, and it has obviously taken the text as a series of discrete phrases which it is unable to connect. But there is light at the end of the machine translation tunnel.

  3. @John Miller – thank you for this. I have looked into this, and the book 3 of Bede’s homilies, containing all this material is in fact spurious. I’ve added a note to the article tho. I appreciate the tip!

    @Paul Chandler – I think verse is a pretty stiff test. Some way to go yet, perhaps.

  4. Back in 2010 when Google Translate launched their Latin tool I tested it with Aeneid VI.268 ‘ibant obscuri sola sub nocte per umbram’ just to see what it did with a not-too-complicated but somewhat tricky bit. The result was “On they went dimly, beneath the lonely night through the gloom”, which sounds beautiful, and so it should, since it’s almost literally the Loeb translation. Clearly the engine had Loeb in its bilingual corpus, found a perfect match and returned the corresponding chunk, which is the sensible thing to do in such cases. No need to try to be clever and improve on the human translation you already have.

    Now in 2022 it produces instead “they were going dark alone under the night’s shadow”, which is several steps back. I suppose that the process has become much more sophisticated and that there are senses in which the new version can be said to be “enhanced”, but if you ask me this is a clear case of “machine unlearning”.

  5. That’s a lovely example! Thank you.

    Of course google were a search engine company, so their results are not from any kind of AI but from searches.

Leave a Reply