3 ms·

Frankly I didn't understand everything in the paper because of the maths/language involved but yes it's a super interesting read. The author has other papers (i

Frankly I didn't understand everything in the paper because of the maths/language involved but yes it's a super interesting read. The author has other papers (including new ones!) on his website: http://www.wasanjyorin.com/ http://www.wasanjyorin.com/ (under section 漢文訓読関係 of course). I don't think that Chinese is more suitable for maths representation per se than other any other natural language. It's the difference between Chinese SVO and Japanese SOV that gave birth to kaeriten and the crazy idea of reading a text in a foreign language in it's own mother tongue by reordering the words. It gives this assembly-like feel to the whole thing, where kanji=memory cells and kaeriten=instructions. My Classical Chinese game is not too strong, but from what I understand the programming language presented is indeed very well crafted.

lioeters 7y ago
Thanks for the link to the author's site - most of it's over my head, but I love how deep he's into this particular topic. Browsing the papers, like 数理漢文学への道 (the road to the "mathematical study of Chinese literature"..?), I see that the reality of working with a natural language is fraught with exceptions and ambiguities, requiring more and more logical rules to be added in order to closely represent the real-life usage of grammar. So I think you're right that Chinese is not necessarily easier than any other language, to diagram this way. It's due to my lack of knowledge, that I imagined Chinese grammar to be more strictly logical than, say, English which is well-known for its quirks and idiosyncracies. In a way, the approach of mathematically representing a natural language reminds me of earlier years of AI research before machine learning - like expert systems and inference engines - to manually build models of knowledge and mental processes. (Well, in the case of this kundoku research, it looks like it's concerned with the syntax structure only, and not about the content/meaning of the words and characters.) > It gives this assembly-like feel to the whole thing, where kanji=memory cells and kaeriten=instructions. Wonderful imagery - I can see why you were working on "a virtual machine for transforming texts (texts are code and data) in kanbun". I didn't get it before, but when you said "texts are code" - now I see that you were feeding kanbun text into a VM that interpreted it as code..!? A somewhat related topic I'm interested in is mathematical representations of music. It feels like a similar endeavor, to build up a set of logical rules to attempt to model the real-life usage of musical "grammar". There are so many exceptions and "illogical" (i.e., strangely dissonant) ways that people express themselves, that will always be "one step ahead" of any set of rules. (On the other hand, there are some fascinating examples of algorithmic music, beyond what humans are able to create.) Anyway, thank you for an introduction into a thought-provoking topic.
- tasogare 7y ago
  > I didn't get it before, but when you said "texts are code" - now I see that you were feeding kanbun text into a VM that interpreted it as code..!? I don't remember the exact details of my thoughts at the time, by I still have the code. I put it online here for you: https://dotnetfiddle.net/Q0kiXU https://dotnetfiddle.net/Q0kiXU As you can see, it works well on small samples but fails for the most complex examples present in the paper. This is because the way I handle ㆑ which is not enough to deal with its real complexity. Edit: reading that https://ipsj.ixsq.nii.ac.jp/ej/?action=repository_uri&item_id=137405&file_id=1&file_no=1 https://ipsj.ixsq.nii.ac.jp/ej/?action=repository_uri&item_i... I thing I was very close. The paper mention three stacks. I'm using 5 in my code, but maybe not in the good way. I don't have too much time to spend on the matter, but I might read a few papers and maybe contact/go meet the author because it's fascinating and nothing is published in an European language.
  - lioeters 7y ago
    Cool, a Kanbun parser and interpreter! That's a really interesting approach. I'm not so familiar with C# but I get the general picture, as I've worked with mathematical expression parser/evaluator before in a couple other programming languages. Indeed, after reading the linked paper, I see how the handling of "㆑" has complex rules. The complete algorithm is beyond my understanding, and I imagine getting it to a practical level would be years of work.. Sounds like it would be fun for you to contact the author, I'm sure he'd appreciate that someone had the courage to dive into the depths of his obsession/expertise in this topic. :)

lioeters 7y ago

Thanks for the link to the author's site - most of it's over my head, but I love how deep he's into this particular topic. Browsing the papers, like 数理漢文学への道 (the road to the "mathematical study of Chinese literature"..?), I see that the reality of working with a natural language is fraught with exceptions and ambiguities, requiring more and more logical rules to be added in order to closely represent the real-life usage of grammar. So I think you're right that Chinese is not necessarily easier than any other language, to diagram this way. It's due to my lack of knowledge, that I imagined Chinese grammar to be more strictly logical than, say, English which is well-known for its quirks and idiosyncracies. In a way, the approach of mathematically representing a natural language reminds me of earlier years of AI research before machine learning - like expert systems and inference engines - to manually build models of knowledge and mental processes. (Well, in the case of this kundoku research, it looks like it's concerned with the syntax structure only, and not about the content/meaning of the words and characters.) > It gives this assembly-like feel to the whole thing, where kanji=memory cells and kaeriten=instructions. Wonderful imagery - I can see why you were working on "a virtual machine for transforming texts (texts are code and data) in kanbun". I didn't get it before, but when you said "texts are code" - now I see that you were feeding kanbun text into a VM that interpreted it as code..!? A somewhat related topic I'm interested in is mathematical representations of music. It feels like a similar endeavor, to build up a set of logical rules to attempt to model the real-life usage of musical "grammar". There are so many exceptions and "illogical" (i.e., strangely dissonant) ways that people express themselves, that will always be "one step ahead" of any set of rules. (On the other hand, there are some fascinating examples of algorithmic music, beyond what humans are able to create.) Anyway, thank you for an introduction into a thought-provoking topic.

tasogare 7y ago
> I didn't get it before, but when you said "texts are code" - now I see that you were feeding kanbun text into a VM that interpreted it as code..!? I don't remember the exact details of my thoughts at the time, by I still have the code. I put it online here for you: https://dotnetfiddle.net/Q0kiXU https://dotnetfiddle.net/Q0kiXU As you can see, it works well on small samples but fails for the most complex examples present in the paper. This is because the way I handle ㆑ which is not enough to deal with its real complexity. Edit: reading that https://ipsj.ixsq.nii.ac.jp/ej/?action=repository_uri&item_id=137405&file_id=1&file_no=1 https://ipsj.ixsq.nii.ac.jp/ej/?action=repository_uri&item_i... I thing I was very close. The paper mention three stacks. I'm using 5 in my code, but maybe not in the good way. I don't have too much time to spend on the matter, but I might read a few papers and maybe contact/go meet the author because it's fascinating and nothing is published in an European language.
- lioeters 7y ago
  Cool, a Kanbun parser and interpreter! That's a really interesting approach. I'm not so familiar with C# but I get the general picture, as I've worked with mathematical expression parser/evaluator before in a couple other programming languages. Indeed, after reading the linked paper, I see how the handling of "㆑" has complex rules. The complete algorithm is beyond my understanding, and I imagine getting it to a practical level would be years of work.. Sounds like it would be fun for you to contact the author, I'm sure he'd appreciate that someone had the courage to dive into the depths of his obsession/expertise in this topic. :)