Episode 2 engl.: Beware of spelling aids and text prediction!

I regularly feel patronised when auto-text predictions are converting my rarer (low-frequent) words into other, more frequent ones. It's not my fault that the respective AI has a limited vocabulary or at least one that differs significantly from mine, is it? Only yesterday, WhatsApp insisted on disimproving my greeting to a friend (a Niki de Saint Phalle aficionado) despite several slow and deliberate typing attempts. My photo of the Nanas in Hannover with the words The Nanas greet you! was consistently changed to “The pineapples greet you!” No, they don't! – Why is there no visible learning effect when I type in Nanas three times in a row? I thought AI was so incredibly adaptive!?

The Nanas by Niki de Saint Phalle on the Leine (Leibnizufer) in Hannover.
Incidentally, my attempt to save this photo as large Nanas was changed to “large mums”...

As a text-focused person, I also falter when this pattern is repeated at the spelling level and capitalisation is used automatically after an ordinal number that I have written with a dot. An example would be a sentence like She 1stly didn't buy anything at all and 2ndly wasn't even out window shopping. The capitalisation applied by the spelling aid here is confusing: “She 1stly Didn't buy anything at all and 2ndly Wasn't even out window shopping.’ – Let's face it: As a German native speaker, capitalisation is relevant to me. – The same thing happens in some programs when you use an abbreviation such as bspw. (beispielsweise, for example) or z.B. (zum Beispiel, for example) and then want to continue writing in lower case. Not every full stop automatically indicates the end of a sentence, even if that is the most common use. These little things can be annoying.

On the other hand, I came across a really funny mistake a few years ago on the train to Leipzig. I wanted to quickly read up on the city and its people on the internet and had curiously clicked on the description of Leipziger Allerlei (a vegetable dish). The title was spelled correctly, but in the body text, ‘Leipziger Allergie’ (Leipzig allergy) had crept in. I have been wondering ever since, what that might actually be.

For anyone who is now curious about the dish, here is a photo: 

A website for tourism in Saxony hides a sneaky disimprovement (photo: 16 October 2021).

I also found it entertaining to be suggested “Alea Aquarium” for Alea Aquarius, the heroine of the young adult novels of the same name. In these, the heroine and her friends whoosh across the expanse of the oceans and campaign for global environmental protection. Definitely not in the dimensions of an aquarium, no matter how big!

My message “The family is growing!” (I had been given a camellia, and I wanted to share that it was growing well, hence The camellia is growing!) was borderline confusing. Sent unnoticed and uncorrected this message triggered astonished speculation about our family planning among the addressees.

A pure reason to laugh on the other hand was a friend's suggestion to a friend with the flu to consume chicken soup. That is, she wanted to suggest it. Because it was changed to corns without being noticed. [engl. corn = dt. Hühnerauge, literally chicken eye]

I am sure there are many other examples of interfering spelling aids and text predictions. Some are funny, some are embarrassing, and some have really bad consequences. Drop me a line if you like, so I can share your experiences here.

Of course, there have been text editing errors long before the computer age. I found one example on a half-timbered house in Lower Saxony. The rhyming Latin inscription on a beam should obviously read: Post lachrima risus post exilium paradisus. Translated: After the tear, laughter; after the exile, paradise. (Here I have pondered the difference between lachrima and lacrima – but I will leave the topic of linguistic variation for another time.)

Peterssches Haus (Oldest half-timbered house in the town of Springe am Deister).

At the latest during the last renovation, the L in exilium (exile, banishment) was no longer correctly painted. Although space was left for it when carving. However, there was so little of it that the unfavourable spatial layout when carving may have abetted a misinterpretation during the renovation. This was possibly exacerbated by a later damage to the beam, in which the lower line of the L was lost. Now the house reads post exiiium paradisus. Was the gilder unable to read Latin or perhaps had an incorrect writing template because the reading error already occurred at the level of placing the order? Or did they decide to leave the damaged beam and to conceal the partial absence of the capital L with a golden I, which can also be a lower-case l? Perhaps the use of a lower-case letter in an inscription otherwise written in capital letters seemed an acceptable evil? Who knows? – (I feel I may be overthinking this. Occam’s razor and all that.)

In the word exilium, there is first an X (also an abbreviation for ten) and then a V (also the Latin for five, here for U). Because there are supposedly numbers written there, it is possible that all five characters were reinterpreted accordingly when they were gilded. The two Is and the narrow L were reinterpreted as three Is and repainted accordingly. Result something like: poste 12 4 m paradisus. Hmmm. Also possible. Do you have a cleverer explanation?

Whatever happened on the beam of the Peterssche Haus, then as now the following truth is valid:

The more data you feed the text-processing instance beforehand, the more accurate the end result will be!

So much for the subject of text prediction glitches. Errors in reading and hearing, as well as slips of the tongue, are completely different but equally fruitful topics. The same goes for translation errors.

But I'll tell you about those another time!

My heartfelt thanks for this high-quality detail photo go to an attentive Springer reader of this blog (received on 05/04/2025):'
It can be seen here that the L in exilium is still partially preserved.