Language Log

Jonathan Silk wonders how this mistranslation from Latin to Dutch in Google Translate occurred the same way in English:

Like so many other good things in this annus horribilis, COVID killed it.

For quite a few years now, I have reported on the national spelling bee (usually in May).  This has been such a dismal year that I didn't make an effort to inquire about what happened with it this spring.  Now, however, as I am preparing a post on Indian feats of memorization, I could not help but wonder about the fate of the 2020 national spelling bee.  Here's what I found out.

"Tough words, little drama, familiar champ in virtual bee", By BEN NUCKOLS, AP (May 29, 2020)

Alison Frankel, "Lexicographer (and Scalia co-author) joins plaintiffs’ team in Facebook TCPA case at SCOTUS", Reuters 10/20/2020:

Can a lexicographer fend off the combined forces of Facebook, the Justice Department and the entire U.S. business lobby at the U.S. Supreme Court?

What if said lexicographer is also the co-author, with Justice Antonin Scalia, of a landmark book about textualism that is cited multiple times in the other side’s briefs?

Bryan Garner – the Black’s Law Dictionary editor, legal writing consultant and, with Justice Scalia, author of Reading Law – has joined the Supreme Court team of Noah Duguid, a Montana man who sued Facebook in 2015 for violating the Telephone Consumer Protection Act. And though he’s only been working with Duguid’s other lawyers for a matter of weeks, Garner’s influence on Duguid’s just-filed merits brief is unmistakable. Who else could so boldly assert that the TCPA’s meaning depends on whether the statute’s “adverbial modifier” applies to just one or both “disjunctive verbs” with a “common object”?

Without taking anything away from the well-deserved kudos for Bryan Garner, I want to underline how odd it is to suggest that without his help, lawyers couldn't be expected to understand simple grammatical concepts like "adverbial modifier", "disjunctive verb", and "common object".

Trending on Weibo, a Chinese microblogging website:

[So as not to give anything away, all syllables are separated and not divided into words.]

Nǐ de huò lā lā lā bù lā lā bù lā duō? Huò lā lā lā bù lā lā bù lā duō yào kàn nǐ de huò lā dé duō bù duō. Rú guǒ lā dé bù duō jiù lā nǐ de lā bù lā duō, rú guǒ lā dé duō jiù bù lā nǐ de lā bù lā duō.

"你的货拉拉拉不拉拉不拉多？货拉拉拉不拉拉不拉多要看你的货拉得多不多。如果拉得不多就拉你的拉不拉多，如果拉得多就不拉你的拉不拉多。"

Google Translate:

"Your cargo pulls, pulls, pulls, pulls, pulls, pulls, pulls, pulls, pulls, pulls, pulls, pulls, pulls, pulls, pulls more? If you pull too much, it won’t pull you.

Before turning the page, if you know Mandarin, try to parse and translate the above sentences.

30 years ago, Don Hindle explored the idea of calculating semantic similarity on the basis of predicate-argument relations in text corpora, and in the context of that work, I remember him noting that we tend to purchase wine but buy beer. He didn't have a lot of evidence for that insight, since he was working with a mere six-million-word corpus of Associated Press news stories, in which the available counts were small:

So for today's lecture on semantics for ling001, I thought I'd check the counts in one of the larger collections available today, as an example of the weaker types of connotational meaning.

This was supposedly Marco Polo's word for Japan.  It has recently come back in vogue for films, games, etc.  It would seem that "Jipangu" (also spelled "Zipangu") is cognate with Jap. Nihonkoku / Nipponkoku, Ch. Rìběnguó 日本國, Kor. Ilbon-guk, Viet. Nhật Bản Quốc , but in none of the Chinese topolects I'm aware of does it sound quite like that.  Certainly it would not work for the southern or other topolects that have an entering tone final -k (or some -t) for the last of the three syllables.  Ditto for Japanese, Korean, and Vietnamese.

Even the Sinitic topolects without an entering tone final don't have the right vowel shape / quality at the end to match the -u of Jipangu.

Maybe Marco Polo got it from Persian, the lingua franca of international diplomacy in his time.  Could it be that the phonotactics of Persian could not tolerate / represent any of the Sinitic topolectal forms of 國 directly but transformed one of them into something that sounded to Marco Polo like -gu?

Did Marco Polo get "Jipangu" from the Mongols?  If so, from whom did the Mongols get it?

Wiktionary entry for 日本國.

Wiktionary entry for 國.

Yesterday my wife challenged me to identify the person in a photo she sent. I decided to cheat, by using Google Image Search — and the results were very strange.

We've posted often about weird AI behavior in Speech-to-Text and Machine Translation and other NLP applications. Image processing has its own litany of weirdness, which is not often a topic here for obvious reasons. But this case does have a linguistic aspect, namely the cited links…

We just had a good discussion about a Sinitic language written with an alphabet:

"The look, feel, and sound of Dungan language" (10/15/20)

Under "Selected readings" below, there are listed additional earlier posts about writing Sinitic languages with Romanization.

One of the major advantages of the alphabet over a morphosyllabic / logographic ideopicto-phonetic writing system like the Sinographic script is that it is very easy to order and find / retrieve the entire lexicon with the former, whereas carrying out these tasks with the latter is toilsome at best and torturesome at worst.  See:

Victor H. Mair, "The Need for an Alphabetically Arranged General Usage Dictionary of Mandarin Chinese: A Review Article of Some Recent Dictionaries and Current Lexicographical Projects", Sino-Platonic Papers, 1 (February, 1986), 1-31 pp.

From John pitchford's Twitter feed (@Johnnypapa64):

World’s worst superhero. pic.twitter.com/gkjElEVW6n

— John pitchford🌹 (@Johnnypapa64) August 1, 2020

Scott Bixby & Asawin Suebsaeng, "The Biden and Trump Shows: It’s Mr. Rogers Vs. ‘Someone’s Crazy Uncle’", The Daily Beast 10/15/2020:

“He didn’t spend the whole time yelling, he didn’t piss himself… so this was as best as we could have hoped for,” said one Trump campaign adviser.

Someone asked me about the "as best as" construction, and I was able to refer them to a 15-year-old post, "Asbestos she can", 12/29/2005.

Here are a couple of YouTube videos by way of example:

Today's xkcd: