15 ms·

Knowledge Representation in Sanskrit and Artificial Intelligence

58 points by hedgehog0 5y ago

canjobear 5y ago
Panini's grammar of Sanskrit is an incredible artifact, all the more so since it was likely composed and transmitted entirely orally at first. It is a completely formalized generative grammar of Sanskrit morphophonemics: https://en.wikipedia.org/wiki/P%C4%81%E1%B9%87ini https://en.wikipedia.org/wiki/P%C4%81%E1%B9%87ini
- jimsimmons 5y ago
  Sadly even Panini’s contributions are always presented in s religious context which is such a shame. The man invented Buckus Naur form 2000 years ago and that’s how we honor him
  - unmole 5y ago
    > Sadly even Panini’s contributions are always presented in s religious context I have never heard of this. Do have any examples?
    - jimsimmons 5y ago
      Look up Maheshwara Sutrani. Apparently Panini heard Shivas damaru and discovered Sanskrit morphology. More religious ones claim he was meditating when God appeared and presented it to him. Most people believe Panini just discovered the amazing properties of Sanskrit. In fact Sanskrit is just any other language except Panini’s grammar which is a rigors and unique formalization of it.
- amznbyebyebye 5y ago
  Panini was a genius. Grew up learning the grammar and it helped me in so many countless ways.
- cupcake-unicorn 5y ago
  Pāṇini has a couple of millennia on Chomsky's language hierarchy! How cool is that?
gantagonist 5y ago
Published: 1985-03-15
lewisjoe 5y ago
For context of Non-Indians: though ancient, there is little evidence Sanskrit ever was a common spoken language among laymen. It was the language primarily used for religious purposes. It was/is a language of high culture and political elites. Coming back to the premise of the paper, the grammar of this language is declared to be sufficiently easy to disambiguify such that it's easier for computers to understand. Just to put it in context, this property is actually a part of many other languages including japanese. Pro tip: Most of japanese grammar can be covered with a handful of BNF grammar! Being an indian I know one other language spoken in the south of India that exhibits this property as well - Tamil. Equally ancient as sanskrit but widely spoken language even as of 2021. Point being, the intention of the paper, though it sounds benign, there is a political context to it which cannot be set aside completely.
- Ballu 5y ago
  wait.. put ref when you are saying " political context to it which cannot be set aside completely.". What political context, are you talking about. I am not linguist or AI expert. But most of your so facts are of no where related to paper as well as authenticity of the work. Check the author as well as date of the paper.
  - lewisjoe 5y ago
    This paper has been repeatedly quoted by the far right of India on how one of the elitist ways of doing something is superior and ahead of time. I saw another comment here which linked to a specific instance of usage. I know the author's intention might be benign. I know the paper dates back to the 80's. But seriously, read through the paper and conclude for yourself if a similar argument can't be made for any of the hundreds of languages in existence today.
    - mistrial9 5y ago
      give credit where credit is due - the rest is bloviation, on all sides
    - zepto 5y ago
srvmshr 5y ago
For a fact, this paper had been quoted and superficially extended several times by Sanskrit scholars I have met in the past. Typical arguments include the 'semantics is suitable for knowledge extraction', to 'replacing English as the modicum of computer instruction set' (supported by claims for disambiguity in grammar) [1,2]. There is no evidence to support these ideas. Like 'lewisjoe', I am from the same subcontinent and share the very same opinion. The reasons are more political signaling and intellectual statement delivery than actual science.[3] A similar claim that I have heard from half baked biologists is that gene constituents ACTG will replace binary encoding in future computing (because ACTG have 4 elements as compared to binary 2, and we are the living proof it works!). While genes were researched in MSR for high density storage (Source: yours truly ex-MSR alumnus working with CS+genetics team), there is no hard evidence a base-4 biological system will inherently perform better than the mathematical base-2. We always had octal and hexadecimal bases. These have their strength and weakness. But from the CPU perspectives, base-2 perfectly encodes HI and LO voltage states and hence the adoption. I wish such researchers took a bit of efforts to investigate their claim's feasibility and the underlying processing structure before making grand statements about computational roadmaps. 1. https://bstrategyhub.com/sanskrit-is-the-best-language-for-artificial-intelligence-says-nasa/amp/ https://bstrategyhub.com/sanskrit-is-the-best-language-for-a... 2.https://scroll.in/article/750526/how-sanskrit-came-to-be-considered-the-most-suitable-language-for-computer-software https://scroll.in/article/750526/how-sanskrit-came-to-be-con... 3.https://www.firstpost.com/india/the-sanskrit-non-controversy-why-it-is-indeed-a-superior-language-1813201.html https://www.firstpost.com/india/the-sanskrit-non-controversy...
aatharuv 5y ago
So short summary: This paper from the 1980's, by a researcher then associated with NASA Ames Research Center basically states that the Sanskrit linguistical and grammatical traditions had mechanisms for describing words and their relations (between the 5th century BC, through the 15th century AD), that didn't make it to Western grammatical traditions until the mid 20th century. These same relations are those that were used in Artificial Intelligence linguistic models, and given that the Paninian grammar could fairly unambiguously represent Sanskrit, it would be possible to map Sanskrit into a programmatic model for natural language processing in computers. Note that through decades of misintepretation in the popular press, this has become, "Sanskrit is the ideal language for programming in computers, or even "NASA has endorsed Sanskrit as the ideal language for programming computers".
- unmole 5y ago
  > even "NASA has endorsed Sanskrit as the ideal language for programming computers". It goes beyond that all the way to "Sanskrit learning is essential for Computer [sic] literacy".
inetsee 5y ago
Another language some people believe would make for better human-computer interaction is Lojban. Unfortunately, despite having been around for decades, it hasn't made much progress in this area.
mark_l_watson 5y ago
I heard a talk many years ago at a AAAI conference on this topic. Really cool that a natural language can be unambiguous.
- srean 5y ago
  You would need to define natural language to engage with this debate. What you will find is that a definition that makes the dialect of sanskrit refered to in the paper 'natural' will include other languages that have similar lack of ambiguity.
xony 5y ago
sanskirt is dead language ..
akomtu 5y ago
What's the meaning of "Sanskrit"? Sun-script?
- unmole 5y ago
  Sanskrit can be translated as cultured, as opposed to Prakrit which is natural.
unmole 5y ago
If you find this interesting, you might also like the book Geek Sublime: The Beauty of Code, the Code of Beauty by Vikram Chandra. It explores connections between linguistics and computation and touches upon Paninian grammar.

58 points by hedgehog0 5y ago

Panini's grammar of Sanskrit is an incredible artifact, all the more so since it was likely composed and transmitted entirely orally at first. It is a completely formalized generative grammar of Sanskrit morphophonemics: https://en.wikipedia.org/wiki/P%C4%81%E1%B9%87ini https://en.wikipedia.org/wiki/P%C4%81%E1%B9%87ini

jimsimmons 5y ago
Sadly even Panini’s contributions are always presented in s religious context which is such a shame. The man invented Buckus Naur form 2000 years ago and that’s how we honor him
- unmole 5y ago
  > Sadly even Panini’s contributions are always presented in s religious context I have never heard of this. Do have any examples?
  - jimsimmons 5y ago
    Look up Maheshwara Sutrani. Apparently Panini heard Shivas damaru and discovered Sanskrit morphology. More religious ones claim he was meditating when God appeared and presented it to him. Most people believe Panini just discovered the amazing properties of Sanskrit. In fact Sanskrit is just any other language except Panini’s grammar which is a rigors and unique formalization of it.
amznbyebyebye 5y ago
Panini was a genius. Grew up learning the grammar and it helped me in so many countless ways.
cupcake-unicorn 5y ago
Pāṇini has a couple of millennia on Chomsky's language hierarchy! How cool is that?

gantagonist 5y ago

Published: 1985-03-15

lewisjoe 5y ago

For context of Non-Indians: though ancient, there is little evidence Sanskrit ever was a common spoken language among laymen. It was the language primarily used for religious purposes. It was/is a language of high culture and political elites. Coming back to the premise of the paper, the grammar of this language is declared to be sufficiently easy to disambiguify such that it's easier for computers to understand. Just to put it in context, this property is actually a part of many other languages including japanese. Pro tip: Most of japanese grammar can be covered with a handful of BNF grammar! Being an indian I know one other language spoken in the south of India that exhibits this property as well - Tamil. Equally ancient as sanskrit but widely spoken language even as of 2021. Point being, the intention of the paper, though it sounds benign, there is a political context to it which cannot be set aside completely.

Ballu 5y ago
wait.. put ref when you are saying " political context to it which cannot be set aside completely.". What political context, are you talking about. I am not linguist or AI expert. But most of your so facts are of no where related to paper as well as authenticity of the work. Check the author as well as date of the paper.
- lewisjoe 5y ago
  This paper has been repeatedly quoted by the far right of India on how one of the elitist ways of doing something is superior and ahead of time. I saw another comment here which linked to a specific instance of usage. I know the author's intention might be benign. I know the paper dates back to the 80's. But seriously, read through the paper and conclude for yourself if a similar argument can't be made for any of the hundreds of languages in existence today.
  - mistrial9 5y ago
    give credit where credit is due - the rest is bloviation, on all sides
  - zepto 5y ago

srvmshr 5y ago

For a fact, this paper had been quoted and superficially extended several times by Sanskrit scholars I have met in the past. Typical arguments include the 'semantics is suitable for knowledge extraction', to 'replacing English as the modicum of computer instruction set' (supported by claims for disambiguity in grammar) [1,2]. There is no evidence to support these ideas. Like 'lewisjoe', I am from the same subcontinent and share the very same opinion. The reasons are more political signaling and intellectual statement delivery than actual science.[3] A similar claim that I have heard from half baked biologists is that gene constituents ACTG will replace binary encoding in future computing (because ACTG have 4 elements as compared to binary 2, and we are the living proof it works!). While genes were researched in MSR for high density storage (Source: yours truly ex-MSR alumnus working with CS+genetics team), there is no hard evidence a base-4 biological system will inherently perform better than the mathematical base-2. We always had octal and hexadecimal bases. These have their strength and weakness. But from the CPU perspectives, base-2 perfectly encodes HI and LO voltage states and hence the adoption. I wish such researchers took a bit of efforts to investigate their claim's feasibility and the underlying processing structure before making grand statements about computational roadmaps. 1. https://bstrategyhub.com/sanskrit-is-the-best-language-for-artificial-intelligence-says-nasa/amp/ https://bstrategyhub.com/sanskrit-is-the-best-language-for-a... 2.https://scroll.in/article/750526/how-sanskrit-came-to-be-considered-the-most-suitable-language-for-computer-software https://scroll.in/article/750526/how-sanskrit-came-to-be-con... 3.https://www.firstpost.com/india/the-sanskrit-non-controversy-why-it-is-indeed-a-superior-language-1813201.html https://www.firstpost.com/india/the-sanskrit-non-controversy...

aatharuv 5y ago

So short summary: This paper from the 1980's, by a researcher then associated with NASA Ames Research Center basically states that the Sanskrit linguistical and grammatical traditions had mechanisms for describing words and their relations (between the 5th century BC, through the 15th century AD), that didn't make it to Western grammatical traditions until the mid 20th century. These same relations are those that were used in Artificial Intelligence linguistic models, and given that the Paninian grammar could fairly unambiguously represent Sanskrit, it would be possible to map Sanskrit into a programmatic model for natural language processing in computers. Note that through decades of misintepretation in the popular press, this has become, "Sanskrit is the ideal language for programming in computers, or even "NASA has endorsed Sanskrit as the ideal language for programming computers".

unmole 5y ago
> even "NASA has endorsed Sanskrit as the ideal language for programming computers". It goes beyond that all the way to "Sanskrit learning is essential for Computer [sic] literacy".

inetsee 5y ago

Another language some people believe would make for better human-computer interaction is Lojban. Unfortunately, despite having been around for decades, it hasn't made much progress in this area.

mark_l_watson 5y ago

I heard a talk many years ago at a AAAI conference on this topic. Really cool that a natural language can be unambiguous.

srean 5y ago
You would need to define natural language to engage with this debate. What you will find is that a definition that makes the dialect of sanskrit refered to in the paper 'natural' will include other languages that have similar lack of ambiguity.

xony 5y ago

sanskirt is dead language ..

akomtu 5y ago

What's the meaning of "Sanskrit"? Sun-script?

unmole 5y ago
Sanskrit can be translated as cultured, as opposed to Prakrit which is natural.

unmole 5y ago

If you find this interesting, you might also like the book Geek Sublime: The Beauty of Code, the Code of Beauty by Vikram Chandra. It explores connections between linguistics and computation and touches upon Paninian grammar.