What is so ‘scientific’ about Sanskrit? | Hacker News Reader

What is so ‘scientific’ about Sanskrit?

smritiweb.com

96 pointsindexerror51 comments

This article only scratches the surface, I think.

One thing that is worth clarifying IMO: There is nothing inherently “scientific” about Sanskrit as a language by itself: instead what's remarkable are

1. the Sanskrit grammatical tradition, as represented especially by Pāṇini. The Sanskrit grammarians paid special attention to language, and their careful study culminated in a grammar by Pāṇini that was centuries ahead of its time, and is still throwing up new insights. The last several generations of modern linguists have read their own ideas into Pāṇini: from William Jones at the start of historical linguistics to de Saussure to Chomsky to Kiparsky. (The last of whom is an expert in both Sanskrit and modern linguistics.) Panini occupies a place in Indian thought equivalent to that of Euclid in Greek, and explicit comparisons have been drawn between the two. (E.g. all students were expected to master them respectively, these were the sciences that all others sought to emulate, ...) (I remember reading somewhere something to the effect that: among achievements of civilizations, the Egyptians have their pyramids and India has Panini's grammar.)

2. the cultural fact that, Pāṇini's grammar having been so exhaustive/complete, it became regarded as the definition of correct Sanskrit, and culturally for centuries everyone who composes in Sanskrit has just chosen to stick to this grammar. This "froze" Sanskrit and gives it a timeless quality: Sanskrit poets composing today use the same language that would have been comprehensible centuries ago; learning this one language opens up literature spanning 2500 years. With other languages that evolve naturally, this is difficult: even Shakespeare's (early) modern English is difficult for many.

TLDR: Sanskrit was formalized very early on.

Well, yes. Thanks. What I was trying to say is that the formalization is actually great:

- it is useful to linguists. (Modern linguistics borrows terms like sandhi and bahuvrihi for example.)

- it is acceptable to language users. Poets have continued to compose in the same language, without finding the formalization artificial.

- it is interesting even as an achievement of the human mind/culture.

A somewhat trivial example, but possibly of interest to the HN audience: the grammar needs to refer to various sets of sounds (like "nasal consonants", "short vowels", "aspirated stops" and so on). To compactly encode these sets, it adopts a curious ordering of the sounds, so that each required set can be represented as a certain substring of that sequence.

It turns out that this ordering can be shown to be optimal: see e.g.

- Economy and the Construction of the Śivasūtras by Paul Kiparsky, Stanford University http://www.stanford.edu/~kiparsky/Papers/siva-t.pdf

- A Mathematical Analysis of Pāṇini’s Śiva sūtras, by Wiebke Petersen, University of Düsseldorf http://user.phil-fak.uni-duesseldorf.de/~petersen/paper/pete...

Thanks for the link to Kiparsky's paper on mAheshvarasutraNi. Starting from the system of grouping and ordering, sandhi rules became straightforward to understand. When I first learned the sutras in school, we were presented with a lot of hand-waving and circular reasoning (sandhi rules- sutra grouping - sandhi rules) about why they were ordered in that seemingly arbitrary manner. It is cool that Kiparsky has solved the inverse problem of figuring out Panini's generative system for the grouping and ordering of syllables!

I am currently studying Panini's Ashtadhyayi (the authoritative source of Sanskrit grammar) and I can tell you from first hand experience what is true and what is not. That Sanskrit is scientific, mathematical, ideal for AI/Comp Sci is gross exaggeration. Panini's work is truly astounding. It is extremely concise and dense. He uses some concepts to explain grammar rules that are normally not used in natural language grammars. For example, he assigns names to certain concepts and then builds on those concepts to create more and more complicated rules. There are rules that encompass other rules. There are rules that prohibit rules declared elsewhere in the book under certain circumstances. You can obtain a verb form by systematically applying 5-6 rules successively. If you know how to apply the rules you don't have to remember verb forms. They can be derived (easier said than done because there are 4000 rules and multiple rules can be applied with plenty exceptions and there are exceptions to exceptions). Panini was far ahead of his time and his work is truly amazing but there is nothing scientific/AI about it. He was a genius who devised a very clever way to succinctly list immensely complicated rules of Sanskrit grammar. Because Sanskrit grammar relies on very precise rules it is very amenable to computational analysis. However this does not mean that it is useful in Comp Sci/ AI as many Sanskrit zealots brag without any proof whatsoever.

I know how the Sanskrit/AI thing started (it's alluded to in the linked article, but not clearly enough).

In 1985, Rick Briggs, a NASA employee and Sanskrit (over)enthusiast, wrote an article in AI magazine [1] pointing out that for knowledge representation, one could use the Navya-Nyaya language (which in his article he called "Shastric Sanskrit"). As this (highly artificial) language (or "dialect of Sanskrit"?) had been created by the Navya Nyaya school precisely for knowledge representation (in some sense), this is not very surprising. A good description of this language is given by Ganeri ([2][3]). I think it's good to publish such articles to show "related work" across cultures (and centuries). There are no clear advantages to using this Navya-Nyaya language over some other formal language though.

In popular culture, the awareness of the existence of this paper morphed from "a particular formal dialect of Sanskrit devised by logicians for knowledge representation happens to be good at knowledge representation which we need in AI" into "Sanskrit is useful for AI" into "Sanskrit is great for computers".

[1]: https://www.aaai.org/ojs/index.php/aimagazine/article/view/4...

[2]: http://www.columbia.edu/itc/mealac/pollock/sks/papers/Ganeri...

[3]: http://www.academia.edu/download/30301777/Logic_Navya-Nyaya_...

Thank you for providing this information. It is amazing how everybody seems to have an opinion about Sanskrit (not talking about you but the media/government etc.) without actually having spent any effort to learn or understand it.

I feel like I have read this exact article before, but instead of Sanskrit it was about Greek. I know this adds nothing of value to the discussion, but I thought it was an odd deja-vu. Does anyone perhaps know which article I am talking about?

Not about mathematics, but Heidegger subscribed to the dumb notion that philosophy could only be done in Greek and German: https://philosophy.stackexchange.com/questions/37972/what-is...

Kind of similar reasoning that Church services can only be done in Latin/Greek. There is a bit of merit to this from that angle in that when you go from a more expressive language to a less expressive language, you inevitably simply cannot translate or express certain concepts.

What makes a language more or less expressive? Do you have scientific evidence for your shocking claim that e.g. English is "less expressive" than Latin?

You are right, different doesn't mean more or less expressive. I was merely going on my experience being a Greek and English speaker, and a Latin student -- there are concepts that simply do not translate over to English, and I have not encountered this hurdle when translating from English to Greek.

Why do you require scientific evidence to support opinions? What's so shocking about claiming English is "less expressive" than Latin? Care to show scientific evidence that suggests that isn't the case?

Please post your scientific references using APA format in all of your comments from now on. Anything less is a disappointment to your high standards.

I took Sanskrit as an undergrad at Princeton as an alternative to Latin. It was hard to absorb all the declensions, to be honest, and still connect to the text. The assumption is that you were eventually going to read Bhagava Gita or some other religious text.

Also, the Sanskrit course got very political, for reasons I could not relate to at all, involving the history of India. The textbook remains on my bookshelf, gathering dust.

This is a problem with a lot of Indian studies[1]. The subjects are so intertwined with religion, culture, mythology (to say nothing about the biases and opinions of the people who would be teaching you) that while on one hand it helps in establishing context, on the other hand it distracts from the subject matter which honestly are deeply interesting independent of their cultural baggage.

[1] In case you're curious, what exactly I mean by Indian studies, some examples are all form of classical Indian music (and that's quite a few ..if you are to consider learning specific musical instruments by themselves as a subject, besides music theory), classical Indian dance forms, Yoga, Ayurveda ...etc

The link "which claims that one of the great attributes of Sanskrit is that the same sentence can have two or more completely different meanings" explains clearly how this happens. Names in Sanskrit are simply descriptive attributes. Things traditionally have many names (which are simply descriptions of different aspects) but that seems unimportant in this context.

I don't see how that interferes with it being used as a programming language, although it doesn't seem particularly useful for the AI problems it was proposed for. There is also no need to carry over the tradition of referring to the same thing with different descriptions in a programming context.

It seems kinda neat that every name is necessarily descriptive as well.

Am I missing something, perhaps the ambiguity goes deeper than that?

It's a language with a long cultured tradition, that was formalized very early on. Of course that early formalization would have an impact, that's what you would expect.

The many layers of meanings is again what you would expect to emerge after a long history of entangling various fields of knowledge, such as philosophy, history, theology, etc, in a dense web with language at its core seen as a primary tool of exploration in these fields (their metaphysics postulated that words have "power" in and of themselves, so language automatically acquired very high status).

Some of these developments are also seen in Europe historically, but not to the same extent. Probably because there was no single unifying language there to play that role (Latin did play that role, but too briefly, and there was no continuity on that level between the ancient Greek tradition and the latinizing Middle Ages - while the whole eastern half of that cultural ecosystem skewed towards a Slavic lingua franca, and even then mostly for religious purposes).

The ability to compose sentences with multiple correct and interrelated meanings made me want to learn it. That sounds fascinating and beautiful, the best possible language for poetry.

But computers hate ambiguity.

Which is why the whole "computer language" meme is probably bullshit.

It's definitely good for poetry, however.

But as I understand it the ambiguity is not in the grammar, but only related to the high prevalence of homonyms.

Speaking about Urdu (among different languages, derived from Sanskrit):

It is one of the languages in which every language, word can be copied into and pronounced to a surprising accuracy. For example, this very post can be written using Urdu syntax and still be pronounced closely to as it would sound in English!

While Sanskrit (and to an extent Hindi) may have some unique properties, I find the syntax (script/alphabet) extremely daunting to understand (for a non Hindi/Sanskrit speaker).

But ultimately, research should continue I guess to see if there is any Sanskrit-derived benefits for the mankind.

I'm lost, what does pronunciation have to do with syntax?

We need more articles like this. The more vocal chest thumping 'patriots' do a great disservice to the rich fascinating history of India by mixing impressive facts with laughable fiction.

I would think if any human-speakable language were promoted for AI uses, it would be the machine parsable language based on formal logic, Lojban. I've thought for a while that this is the perfect halfway-point for human-computer communication.

https://mw.lojban.org/papri/Lojban

What makes a language scientific? Apparently, modern science is born in the land where people don't bother to persuade others that their language, by nature, is more scientific than the rest.

The push of such argument is inherently nationalistic than 'scientific', ironically.

Reason and truth are so refreshing to the ear! I commend you for being objective on this topic.

the recursive infinite nature of it ?

This is exactly the sort of vague CS-flavored hype that the article is criticizing. You've said nothing that distinguishes Sanskrit from any other language.

What does that mean, exactly?

Not much, but old languages seem ad-hoc in nature, see measurement units before the metric system, or counting subsystems (4x20 in French for instance). It was a discontinuous aggregate of pragmatic descriptions. Indians very early into fully regular and deep nesting (based on the few thin bits I've read).

Panini's work is considered to be the precursor of Backus-Naur form.[1]

[1] https://en.wikipedia.org/wiki/Backus%E2%80%93Naur_form#Histo...

In fact it was generative. In that, you could not speak the language for 1000 years and still generate all valid strings that can be thought of as members of the language

Why isn't that trivially true of all languages? You can generate infinitely many valid strings of arbitrary length in any grammar. How's that special about Sanskrit?

I showed this to my wife. She called the article and the author a troll. I'm looking forward to her explanation why.

Interesting stuff. The Korean language is considered one of the most scientific language on the planet. I'm puzzled why that NASA guy would push Sanskrit to be an actual application language. In my opinion, an HN'er should be able to read and write Korean in about an hour. I don't know if that's true for Sanskrit.

Taking a look at Sanskrit, it looks very cool. I think it's aesthetically better looking than Korean alphabet.

http://www.omniglot.com/images/writing/sanskrit_cons.gif

I challenge that most could learn to read and write Korean within 15 minutes [0]. At most, stumbling over some of the vowels. It really is a marvelously simple language. Of course - reading it at speed will still take more practice but I feel one could become quite proficient at reading (without any comprehension of what they are reading) within a few weeks of practice.

[0] http://www.ryanestrada.com/learntoreadkoreanin15minutes/

If ^ makes the 's' sound in summit, why is ^| = "she" and not "see"?

TL;DR: Palatalization.

https://en.wikipedia.org/wiki/Palatalization_%28sound_change...

https://en.wikipedia.org/wiki/Korean_phonology#Vowel_assimil...

That "ee" sound is known in IPA [0] as /i/. That sound is also the vowel counterpart to the semi-vowel /j/ [1, 2]. Within languages, /j/ has a tendency to palatize sounds. In the case of Korean, this has resulted in /s/ being conditioned to /ɕ/ before /i/, where /ɕ/ is the alveolo-palatal version of /s/. (This sound is technically different than "she", which is /ʃi/.)

This process of moving where a sound is pronounced is actually known as "assimilation", but this particular case of assimilation is via a palatal sound. That's why there's actually two palatalization terms in linguistics -- the one I linked above and this one: https://en.wikipedia.org/wiki/Palatalization_%28phonetics%29

[0] https://en.wikipedia.org/wiki/International_Phonetic_Alphabe...

[1] English "y" sound, as in "you" /ju/.

[2] https://en.wikipedia.org/wiki/Semivowel

>one could become quite proficient at reading (without any comprehension of what they are reading)

If you can't comprehend what you're reading, you're not reading. Learning the Korean alphabet is not learning Korean.

I disagree entirely. You can read without comprehension - hell I can read my native language without comprehending what I'm actually reading. I'm still reading when I do so. What else would you call it when reading a heavily technical document you don't fully understand? You may not comprehend what the words as a whole mean but you're still reading.

FWIW, only one dictionary definition I can find includes "comprehend" in the definition and it still falls under other definitions such as "speaking aloud written or printed material to other people".

I never claimed it was. My post was very specifically about learning to read and write Hangul.

Korean and Hangul are different. If you know the Latin alphabet that doesn't mean you know how to read English. It means you know how to read the Latin alphabet.

Of course, but can you do the same with Arabic in the same time span? I think not.

With Hangul, you can start writing your own name, lot of English words in Korean which is understood in Korea. Even Konglish (Phonetically typed Korean in English) is understood to a certain degree.

Anything beyond that, you will need to learn the Korean words, but it gets you a lot farther than most languages.

No I don't have scientific evidence to support this but from anecdotal experiences. Somebody wants to write a paper on it let me know, I'll give you 0.1 bitcoin.

You aren't learning Korean. You are learning Hangul. You don't just need to learn 'the Korean words'. You need to learn Korean lexicon, Korean syntax, Korean morphology, Korean idiom, Korean pronunciation, etc. etc. Those are the actually difficult parts of language.

Learning the Latin alphabet doesn't mean you have learnt to read English.

And sure of course you can learn the Arabic alphabet quickly. It's a lot simpler than it looks.

> The Korean language is considered one of the most scientific language on the planet.

It's considered that by lay people, largely due to Korean patriotism, not by real linguists (can you provide a source?)

> an HN'er should be able to read and write Korean in about an hour.

So, you're talking about the alphabet used to write the language, not the language itself. At best you've proven that Korean has a highly regular writing system. But I'd argue that the average HNer could learn to read, for example, Spanish in about two minutes.

> It's considered that by lay people, largely due to Korean patriotism, not by real linguists (can you provide a source?)

I didn't say the MOST, it's among the more scientific languages. I'm not pushing for Korean patriotism, that's just your own assumptions.

> So, you're talking about the alphabet used to write the language, not the language itself. At best you've proven that Korean has a highly regular writing system. But I'd argue that the average HNer could learn to read, for example, Spanish in about two minutes.

Hangul is the writing system that is uniquely different from a latin based system like Spanish. Your argument is moot.

So, can you give me a scientific/academic source that says anything about "how scientific" Korean is?

> Hangul is the writing system that is uniquely different from a latin based system like Spanish.

How? They're just different alphabets. The only unique feature of Hangul is that all the letters making up a syllable are arranged in one square -- is that what you're referring to as making it uniquely different? If not, what?

> So, can you give me a scientific/academic source that says anything about "how scientific" Korean is?

https://scholar.google.ca/scholar?q=korean+language&btnG=&hl...

> The only unique feature of Hangul is that all the letters making up a syllable are arranged in one square

That's one feature that makes it superior to say Chinese. Chinese character is a bitch to learn and it's been bastardized by the communist party to increase literacy.

Korean alphabet, not language.

Ha, this is exactly the other language that came to mind when I saw this article. I love Hangul, but a lot of its "scientific"-ness comes from it being so recent that the spoken pronunciation hasn't drifted from the written form. There just a few rules for changing sounds, and they're quite regular.

Unfortunately the people in charge of the Revised Romanization decided that these rules should be applied to the Roman alphabet as well, so we have execrable constructions like "Daegu" or "bibimbab" which can't be pronounced from English spelling rules.

You really should be more precise. I'm sure you meant "orthography" in every usage of "language". And yes, Hangul is a relatively "shallow" or "transparent" orthography -- though it is becoming less so over time. But so is Spanish's use of the Latin alphabet, Japanese hiragana and katakana, and yes Devanagari.

https://en.wikipedia.org/wiki/Orthographic_depth

Reply on news.ycombinator.com