Language Log

People, including serious linguists, are beginning to wonder:

John McWhorter, "Are translation apps making the learning of foreign languages obsolete?", NYT 7/25/2023

I remember a time, not too long ago, when John was making a serious effort to learn Mandarin, because he often asked me cogent questions about the language and wanted to know the best methods for learning it.

…

What we can learn from the Tower of Babel

In Europe, nine out of 10 students study a foreign language. In the United States, only one in five do. Between 1997 and 2008, the number of American middle schools offering foreign languages dropped from 75 percent to 58 percent. Between 2009 and 2013, one American college closed its foreign language program; between 2013 and 2017, 651 others did the same.

At first glance, these statistics look like a tragedy. But I am starting to harbor the odd opinion that maybe they are not. What is changing my mind is technology.

This short video claims that these two men speak perfect Korean with a Pyeongyang accent.

Or, to put it another way, in the words of Douglas Hofstadter,

Learn a Foreign Language Before It’s Too Late

AI translators may seem wondrous but they also erode a major part of what it is to be human.

The Atlantic (7/13/23)

Hofstadter recounts how he spent years of painstaking, hard labor learning more than half a dozen foreign languages, though he never came close to mastering any of them except French and Italian.

But today we have Google Translate. Today we have DeepL. Today we have ChatGPT—and so on. There’s no need for me to list all the powerful technologies that allow anyone today—a monolingual American, say, who has never devoted a single moment to learning, say, Chinese—to write fluent passages in Chinese. Today it’s a piece of cake to send an email in a tongue you don’t know a word of. You just click on “Translate” and presto! There it is! Or at least, there it is, in a certain sense. Assuming that there are no egregious translational blunders (which there often still are), what you are sending off is slick but soulless text.

I promised that I would tell the story of how five old, white men persuaded me to begin the study of Asian languages two years after I was out of college.  Here it is.

When I graduated from Dartmouth in 1965, I joined the Peace Corps for two years in Nepal.  Although I contracted fifteen diseases, some quite serious, lost fifty pounds, and had three nearly deadly trail accidents, the experience was transformative.

I was an English major in college and wrote an undergraduate thesis on Chaucer's "Troilus and Criseyde".  At the end of my Peace Corps service, I still wanted to study for a PhD on Chaucer.  So, among other applications to graduate school and for funding, I applied for a Woodrow Wilson fellowship.  In those days (1967), that was a very prestigious prize.

As I am about to deliver a keynote address to an international conference on Chinese language pedagogy, I receive news of this new LLM that knocks my socks off:

InternLM is a multilingual large language model jointly developed by Shanghai AI Lab and SenseTime (with equal contribution), in collaboration with the Chinese University of Hong Kong, Fudan University, and Shanghai Jiaotong University.

Technical report: [PDF]

Note: Please right click the link above to directly download the PDF file.

Abstract

We present InternLM, a multilingual foundational language model with 104B parameters. InternLM is pre-trained on a large corpora with 1.6T tokens with a multi-phase progressive process, and then fine-tuned to align with human preferences. We also developed a training system called Uniscale-LLM for efficient large language model training. The evaluation on a number of benchmarks shows that InternLM achieves state-of-the-art performance in multiple aspects, including knowledge understanding, reading comprehension, mathematics, and coding. With such well-rounded capabilities, InternLM achieves outstanding performances on comprehensive exams, including MMLU, AGIEval, C-Eval and GAOKAO-Bench, without resorting to external tools. On these benchmarks, InternLM not only significantly outperforms open-source models, but also obtains superior performance compared to ChatGPT. Also, InternLM demonstrates excellent capability of understanding Chinese language and Chinese culture, which makes it a suitable foundation model to support Chinese-oriented language applications. This manuscript gives a detailed study of our results, with benchmarks and examples across a diverse set of knowledge domains and tasks.

As we have pointed out countless times on Language Log, if one wishes to learn a Sinitic language, one can concentrate on the characters (writing system), one can rely exclusively on romanization or other phoneticization, or one can devise various means for combining the two approaches.  Here is a clever, fun method for learning Cantonese that tackles the problem head on.

Hongkonger creates colourful Cantonese font to foster language learning

Jon Chui’s new font shows coloured, context-sensitive jyutping for Chinese text. He created it as his partner “had a hard time with the tones” when learning Cantonese.

Mandy Cheng, Hong Kong Free Press (5/16/23)

Jon Chui "has created a new Cantonese font, which combines over 8,000 characters with colourful, Romanised pronunciation guides in order to foster language learning and teaching."

Cantonese Font. Photo: Jon Chiu.

Registration is open for Artificial Intelligence in Language Education

Two years ago, during the middle of lockdown when we had to teach all of our courses via zoom, one student was conspicuously superior to all the other dozen or so students in my first-year Literary Sinitic / Classical Chinese (LS/CC) class.  She was clearly an innately smart student, but in addition she seemed to possess a special knack for grasping the grammar, structure, and meaning of the texts we read day after day.  When it came to parsing a particularly difficult passage, she was consistently the one who could figure it out fastest and most accurately.  I had no idea to what particular talent or prior training her excellence could be attributed.

I should mention that this student was from China, as were two-thirds of the others.  Only one-third of the class were from other countries.  I should note, parenthetically, that by and large the more languages a student knows well when he or she takes LS/CC, the better she or he tends to perform in my class.  For example, one of the best students in recent years was a Mexican whose native tongue is Spanish and who is advanced in Korean.  I let him pronounce the texts in Korean.

The following note and video were sent to me by Bill Benzon.  The video is too long (19:15) to make as the main content of this post, but it is captivating, and I warmly recommend it to anyone who is interested in the topics it covers, especially English language learning by Japanese.

The video features Moe, the former geisha who successfully transitioned to Kimono Mom.  Among her interlocutors is a woman from Brazil who has a lot of interesting things to say about Portuguese (especially how different the grammar is from English).  Aside from discoursing on language teaching and learning, Moe is very good at talking about food and cooking for her little family, so if you like that sort of thing, hop on her channel (see below) and you will have many popular videos to choose from.

One of my favorite books for everyday living is Irma S. Rombauer's Joy of Cooking.  The author's cheerful approach to her craft in the kitchen is similar to my jubilant upāya उपाय ("expedient pedagogical means; skill-in-means; skillful means" > fāngbiàn 方便 ["convenient"]) in the classroom.

In my classes, especially Introduction to Literary Sinitic / Classical Chinese (LS/CC), we don't just read through texts with the aid of vocabularies, commentaries, annotations, and grammar notes.  We live the texts, act them out, draw them on the board, debate them, chant them, analyze them, get at their profound philosophical significance, plumb their esthetic depths.

In learning Mandarin, after words for "hello", "how are you?", "goodbye", "yes", "no", "thank you", "where", and a couple dozen other basic expressions, one graduates to the next level of linguistic subtlety by learning "polite talk" (kèqì huà 客氣話) such as "bù hǎoyìsi 不好意思" and "bù gǎndāng 不敢當".

Both are well-nigh untranslatable, at least not with a single English term, since their meanings are so context sensitive.

From the Facebook page of a friend:

[This is a guest post by Paul Shore.]

    The 2022 book Kingdom of Characters by Yale professor Jing Tsu is currently #51,777 in Amazon's sales ranking.  (The label "Best Seller" on the Amazon search-results listing for it incorporates the amusing mouseover qualification "in [the subject of] Unicode Encoding Standard".)  I haven't read the book yet:  the Arlington, Virginia library system's four copies have a wait list, and so I have a used copy coming to me in the mail.  What I have experienced, though, is a fifty-minute National Public Radio program from their podcast / broadcast series Throughline, entitled "The Characters That Built China", that's a partial summary of the material in the book, a summary that was made with major cooperation from Jing Tsu herself, with numerous recorded remarks by her alternating with remarks by the two hosts:  https://www.npr.org/podcasts/510333/throughline (scroll down to the May 26^th episode).  Based on what's conveyed in this podcast / broadcast episode, I think many people on Language Log and elsewhere who care about fostering a proper understanding of human language among the general public might agree that that ranking of 51,777 is still several million too high.  But while the influence of the book's ill-informed, misleading statements about language was until a few days ago mostly confined to those individuals who'd taken the trouble to get hold of a copy of the book or had taken the trouble to listen to the Throughline episode as a podcast (it was presumably released as such on its official date of May 26th), with the recent broadcasting of the episode on NPR proper those nocive ideas have now been splashed out over the national airwaves.  And since NPR listeners typically have their ears "open like a greedy shark, to catch the tunings of a voice [supposedly] divine" (Keats), this program seems likely to inflict an unusually high amount of damage on public knowledge of linguistics.