Sanskrit has 1800 odd verbs in its list of verbs which are conjugated in 10 tenses and moods with 3 different persons (1st, 2nd and 3rd persons) and 3 different numbers (singular, dual and plural). There's additionally active voice and middle voice in which some 30% of the verbs can be conjugated. That's a bit like 1800x10x3*3 = 162,000 odd forms. You can add the dual voice forms to these.
The heart of Sanskrit grammar (articulated by Panini) provides for rule (i.e., sutra) based generation of these 162,00+ forms. However the grammar only has some 4,000 rules (sutras). So Chomsky and others have called it a generative grammar. In addition, the grammar also covers other things like nouns (whose forms i.e., declensions, can all be derived by applying rules of grammar), nominal compounds and various other grammatical categories.
When you derive forms, you have to apply rules. At times, multiple rules are simultaneously applicable. So there's this problem of choice i.e., which is the right rule to choose. Remember the end result is already known from spoken Sanskrit. So the rule that leads you to the right form is the one to be chosen e.g., the plural of 'ox' is 'oxen', not 'oxes'. You are to end up with 'oxen' at the end of the derivational process.
So how do you choose the rule - there are various interpretive devices, some are heuristic (like if the penultimate vowel is like this, do this - Latin students must already start to like this, but there is no antepenult of significance in Sanskrit heh heh), some are phonological (if there's a short vowel at the end of a word, drop it), some are morphological (for instance, the rule that applies to the inside of the word has precedence over the one that operates on the word boundary), or that the particular rule supersedes the general rule etc.
In all instances when you apply interpretive devices to choose a rule, you think like a judge, not a programmer. This is what is the essential difference between Sanskrit grammar's processes and what you find in a program. That's why 'Sanskrit is a programming language' bozos have not created anything of value up until now. Nor will they, in most Sanskritists' view.
Adverting to this thesis and its Eureka moment (I want to kill myself), it takes the view that a central interpretive device in traditional Sanskrit grammar has been wrongly understood and that all the Sanskrit grammarians of some 24 odd centuries were totally out to lunch.
Rajpopat has in his thesis a handful of trivially simple examples, mainly nouns, where his 'correct' interpretation works. Nobody has applied his interpretive device to all the nouns. His method will reduce false positives in a small amount of known examples but would generate myriad false negatives everywhere else. As to the verbal system in Sanskrit, it's so complicated that the derivational processes really require a legal mind. Wisely, Rajpopat sticks to mainly nominal forms in his thesis. As a trivial example, his method may generate 'speeched' as the participial form 'speak' rather than 'spoken' or may generate even 'spought', analogous to 'sought' as the participial form of 'seek'.
I still feel however that his method, after much refinement, may have some utility for machine generation of certain derivations of verbal and nominal forms because it might have a simpler heuristic for conflict resolution in the event of rule conflict. But remember the traditional grammar does it already and has been doing it since Panini wrote his grammar :)
The rest of it is just baseless Cambridge hubris aided by some real high class publicity blitz of the type Cambridge Analytica would have been proud of. Needless to say, Indian media have swallowed it all :) Indians, in general, know no more Sanskrit than a plumber in Chilicothe, OH knows Mycenaean Greek.
The heart of Sanskrit grammar (articulated by Panini) provides for rule (i.e., sutra) based generation of these 162,00+ forms. However the grammar only has some 4,000 rules (sutras). So Chomsky and others have called it a generative grammar. In addition, the grammar also covers other things like nouns (whose forms i.e., declensions, can all be derived by applying rules of grammar), nominal compounds and various other grammatical categories.
When you derive forms, you have to apply rules. At times, multiple rules are simultaneously applicable. So there's this problem of choice i.e., which is the right rule to choose. Remember the end result is already known from spoken Sanskrit. So the rule that leads you to the right form is the one to be chosen e.g., the plural of 'ox' is 'oxen', not 'oxes'. You are to end up with 'oxen' at the end of the derivational process.
So how do you choose the rule - there are various interpretive devices, some are heuristic (like if the penultimate vowel is like this, do this - Latin students must already start to like this, but there is no antepenult of significance in Sanskrit heh heh), some are phonological (if there's a short vowel at the end of a word, drop it), some are morphological (for instance, the rule that applies to the inside of the word has precedence over the one that operates on the word boundary), or that the particular rule supersedes the general rule etc.
In all instances when you apply interpretive devices to choose a rule, you think like a judge, not a programmer. This is what is the essential difference between Sanskrit grammar's processes and what you find in a program. That's why 'Sanskrit is a programming language' bozos have not created anything of value up until now. Nor will they, in most Sanskritists' view.
Adverting to this thesis and its Eureka moment (I want to kill myself), it takes the view that a central interpretive device in traditional Sanskrit grammar has been wrongly understood and that all the Sanskrit grammarians of some 24 odd centuries were totally out to lunch.
Rajpopat has in his thesis a handful of trivially simple examples, mainly nouns, where his 'correct' interpretation works. Nobody has applied his interpretive device to all the nouns. His method will reduce false positives in a small amount of known examples but would generate myriad false negatives everywhere else. As to the verbal system in Sanskrit, it's so complicated that the derivational processes really require a legal mind. Wisely, Rajpopat sticks to mainly nominal forms in his thesis. As a trivial example, his method may generate 'speeched' as the participial form 'speak' rather than 'spoken' or may generate even 'spought', analogous to 'sought' as the participial form of 'seek'.
I still feel however that his method, after much refinement, may have some utility for machine generation of certain derivations of verbal and nominal forms because it might have a simpler heuristic for conflict resolution in the event of rule conflict. But remember the traditional grammar does it already and has been doing it since Panini wrote his grammar :)
The rest of it is just baseless Cambridge hubris aided by some real high class publicity blitz of the type Cambridge Analytica would have been proud of. Needless to say, Indian media have swallowed it all :) Indians, in general, know no more Sanskrit than a plumber in Chilicothe, OH knows Mycenaean Greek.
Just my 2c.