HSK Manga: Turning Manga into Beginner-Level Chinese
2 points by briankung
2 points by briankung
Biasing individual tokens seems like a clever approach but I am suspicious that you can produce comprehensible graded readers just translating a single panel at a time.
As someone who has read a lot of Chinese graded readers, I can tell you that a lot of what makes a graded reader good is what they choose to repeat and what they choose to skip. Good graded readers repeatedly re-emphasize the same concepts, grammar patterns, and words consistently throughout the book to give you varied exposure to them, while at the same time skipping large, non-essential chunks of the story in order to make it not too boring.
I don't think an LLM translating panel by panel is capable of making these sort of decisions.
Good graded readers repeatedly re-emphasize the same concepts, grammar patterns, and words consistently throughout the book to give you varied exposure to them, while at the same time skipping large, non-essential chunks of the story in order to make it not too boring.
I have never used a graded reader; however, I suspect you're completely right about this.
You can get a similar benefit from reading the front page of couple newspapers every day for a few weeks, especially focusing on articles related to some major, ongoing event. Every article will orbit around the same vocabulary and phraseology, but with a certain controlled variation as the story unfolds.
Biasing individual tokens seems like a clever approach but I am suspicious that you can produce comprehensible graded readers just translating a single panel at a time.
I don't think an LLM translating panel by panel is capable of making these sort of decisions.
If you look at the sample chapter they provide, you can see other flaws in this particular approach. Not only does it replace text even if there was no actual change to the wording, but it makes some alterations that seem wholly unnecessary, and some of the alterations are extremely clunky, just to avoid a single out-of-variance terminology.
I am also skeptical that the effort to replace on the text within the image was necessary at all. It's definitely slick, but I think a side-by-side simplification would have been wholly sufficient. (I think this approach would have benefited from more instructional design, but I don't think that was the real purpose of this effort…)
I can understand the appeal of bringing your own source materials and having instructional materials created from them, but I wonder if it's possible to create a good graded reader (by hand or by LLM,) under these circumstances. (I'm also not sure why one needs so much choice in graded readers—is The Little Prince not enough anymore?)
I think that these kinds of pet projects were more beneficial where the majority of the work was done by the author.
Obviously, having an LLM generate a bunch of code on your behalf makes it more likely you will be able to complete something that you can show others, but there seems to be a dilemma in that the better the model is at generating output, the less one is exposed to their own creation. As a result, it is possible to create things that are polished but ultimately not particularly meaningful or useful.
To be clear, even if the vast majority of the work was performed by the LLM, I think it's impressive for a high school student to have created anything at all. But, to be clear, I also am skeptical that the result, that this tool, is actually very useful or interesting for learning.
In my life, I have found that the process of building something by hand draws me into the creation, and it is often only after significant effort that I realise that I shouldn't be trying to build the thing in the first place. Were I to be able to build the thing without this effort and without this friction, it seems I would be robbed of this realisation.
In the case of this particular project, the home page claims “learning Chinese is hard,” and the linked page mentions that “writing [graded readers] by hand doesn't scale” and “engaging graded text is hard to come by.” None of this is true.
The post and the website seem to focus on preparing for the HSK 3 level. While I am not familiar with this exam, as I understand it, the level of proficiency it measures is incredibly low. I suspect that traditional (low-volume) instructional materials would be wholly adequate to achieve this level. In fact, I suspect that traditional materials would likely to be sufficient for the entire set of HSK examinations, and additional materials would be necessary only for developing proficiency beyond these. However, at that level, it's probably the case that one might be able to rely on materials designed for young age native speakers.
Again, it is impressive and laudable to see any kind of output generated by any means by a high school student, but I suspect a project like this is primarily valuable in what building it cultivates in the creator. If mostly everything was LLM-generated, what's left?
I will say as a Chinese language learner that this kind of site is actually very compelling. The texts I was using in college still had dialogues where they would call each other tóngzhì 同志, or "comrade." I've read multiple accounts of people getting deeper into Chinese through dònghuà 动画 (comics, basically Chinese manga), and the HSK exams are a real barrier to entry into Chinese universities and companies. Traditional instructional materials may work, but more comprehensible input is always welcome to drive motivation. Also, Chinese is a category 5 language as classified by the US Foreign Service Institute, putting it in the hardest category of languages for native English speakers to learn.
I agree that it's highly likely that hand building the site would likely cement the techniques more, though.
as a Chinese language learner that this kind of site is actually very compelling
I don't think it's possible to discuss whether or not instructional materials will be well-received by (or even effective for…) an individual learner. There's too much variety in experiences, so we can only speak in generalities.
For Chinese language learning, there are many new apps and YouTube channels and learning methodologies. I would suspect extremely few of them are actually useful or provide material advantage over traditional materials and techniques, but that isn't to say they are actively harmful.
The texts I was using in college still had dialogues where they would call each other tóngzhì 同志, or "comrade."
This is probably Practical Chinese Reader, which was published in the early eighties and was fairly popular in English-language instruction of Chinese. These materials were popular for a few reasons, but more up-to-date materials (including a revised version of this textbook) have been available for many years.
I think many have pointed out how dated these materials are, but, paired with in-person instruction, I don't think it really mattered much at all. I doubt any instructors failed to provide supplementary guidance on where vocabulary choices have changed (or how regional vocabulary differs.) The vast majority of the lessons are still relevant, so I doubt this impeded any actual student learning.
(For reference, I believe the first book of this series ends somewhere around being able to say: “My mother called me on the phone at 10 AM. I was happy. She said she is busy at work.” This is a very introductory level, and while it's not easy to create beginner instructional materials, I think fairly superficial issues like this are not a real barrier for most learners.)
HSK exams are a real barrier to entry into Chinese universities and companies
I don't know much about this, but I suspect the top HSK-levels mark the very earliest point one might be able to begin pursuing a native-language (degree-granting) undergraduate programme in-country, presuming that the student will have the ensuing four years to continue developing necessary language skills.
There are HSK 3 practice exams available online, and, at a glance, the level is very introductory: e.g., “‘Chang’ is my last name; ‘Hsueh’ is my name. We put the surname at the front and the name at the back. That's different than your country.”—“Chinese names are …” ⓐ simple ⓑ usually long ⓒ surname first”
I'm not sure this is a high-enough level to really require any special instructional materials or extraördinary methodologies.
Chinese is a category 5 language as classified by the US Foreign Service Institute, putting it in the hardest category of languages for native English speakers to learn.
Yes, I think this means it is difficult in the same sense that learning Algebra or learning basic Python or learning to tie one's shoes is difficult. There are certainly people who will struggle, and some may repeatedly fail to make progress. But the failure of most English-speaking Chinese learners is, I think wholly unrelated to exaggerated claims of its difficulty. In fact, I suspect these exaggerations significantly worsen the effect of the preposterously low standards these these learners are put to, and (probably coupled with their low inherent motivation,) this is a large contributor to their struggles.
Chinese is a category 5 language as classified by the US Foreign Service Institute, putting it in the hardest category of languages for native English speakers to learn.
Yes, I think this means it is difficult in the same sense that learning Algebra or learning basic Python or learning to tie one's shoes is difficult.
I think you're trivializing the difficulty here. The categorization is based on linguistic distance. There are very few shared cognates, whereas it's relatively easy for English speakers to guess at the meanings of Spanish, and the lexical similarity of, e.g., Italian and French are as close as other languages' dialects are to each other.
I'm not sure this is a high-enough level to really require any special instructional materials or extraördinary methodologies.
I don't think this is a novel instructional method, but rather a cowpath for those who wouldn't otherwise dedicate the hours to strict study.
...either way, I thought the OCR and rendering techniques were more interesting for this particular crowd, but the Chinese language pedagogy seems to have taken a life of its own.
Chinese is a category 5 language as classified by the US Foreign Service Institute, putting it in the hardest category of languages for native English speakers to learn.
Yes, I think this means it is difficult in the same sense that learning Algebra or learning basic Python or learning to tie one's shoes is difficult.
I think you're trivializing the difficulty here. The categorization is based on linguistic distance.
I think FSI categorisation is about relative difficulty: it is, indeed, probably harder to learn Chinese than it is to learn Spanish for an English-language speaker, and, in the aggregate, this means that learning Chinese may require more instructional time to reach an equivalent level of proficiency. It is clearly a very broad organisational tool that is useful when managing large populations of students. I don't think it's particularly relevant or useful in the way it is used to cast Chinese as being hard to learn.
Algebra is a year-long course for many high school students, and many of those students struggle greatly with the subject. Students routinely fail to pass their Algebra classes. Some even fail after repeated attempts. There are many reasons that this happens, and sometimes it is the fault of the instruction, sometimes it is the fault of the student, and sometimes it is related to the inherent difficult of the material itself. However, this course is considered foundational and rudimentary, in that it is expected that the majority of students will be able to learn the material with no need to expend extraördinary effort. It is, for most students (and at population scale,) not considered a unique or even particularly notable achievement to have passed this class.
TOCFL provides similar aggregate, population-level guidance on the instructional hours necessary (in- and out-of-country) to achieve each of its proficiency levels. As we can see, the expected required effort is not low and is higher than what might be expected for other languages. However, it is not only quite achievable, but I think, in practice, we routinely see learners achieve high degrees of proficiency. In this sense, learning Chinese is not easy, in the sense that not everyone can do it, and some who try will fail, but the same could be said of tying one's shoelaces.
In other words, learning Chinese is simply not particularly difficult in the way that demands or warrants extraördinary effort, learning materials, approaches or praise, and, in fact, I assert that it is the exaggeration of its difficulty that creates one of the greater barriers for English-speaking learners. That this exaggeration leads to a proliferation of dubious educational materials, approaches, tools, advice is an understandable consequence. And don't even get me started on the YouTube channels (almost every single “American shocks people by speaking flawless Chinese” video I have been unfortunate enough to see demonstrates ability barely higher than a typical B- grade-average, US-state-university university-sophomore…)