Universal Vedic Multi-Script Transliteration

Transliterating Sanskrit between scripts is often treated as a simple character-by-character string replacement problem. For Vedic Sanskrit, this simplistic approach causes immediate text corruption.

vyasa-lipi implements an Intermediate Representation (AST) architecture designed around the following linguistic invariants:

1. Lossless Round-Trip Guarantee: Transliterating A → B → A must produce byte-identical text for any pair of supported scripts.
2. Suprasegmental Pitch Accent (Svara) Fidelity: Vedic tones (Udātta, Anudātta, Svarita, DīrghaSvarita) must be faithfully mapped across scripts.
3. Strict Roman Separation: ISO 15919 and IAST are distinct linguistic standards and must never be conflated.
4. Phonological Expressiveness: Target scripts must preserve all four stop registers (k, kh, g, gh) to prevent semantic collapse.

---

1. The Brahmic Abugida Architecture

All major indigenous scripts of South Asia (Devanagari, Telugu, Kannada, Grantha, Malayalam, Bengali) belong to the Brahmic script family. Unlike alphabets (where consonants and vowels have independent letters) or abjads (where vowels are omitted), Brahmic scripts are abugidas:

Consonant Base (inherent short 'a')
        ├── With Dependent Vowel Mark (Mātrā):  क + ो  ⟶  को (ko)
        ├── With Virāma (Halanta):               क + ्  ⟶  क् (k)
        └── In Consonant Cluster (Saṁyoga):     क + ् + ष + े  ⟶  क्षे (kṣe)

The Akṣara Token Intermediate Representation

vyasa-lipi parses all scripts into a structured phonemic AST rather than executing regex chains:

pub struct AksharaToken {
    pub initial_consonants: Vec<Consonant>,
    pub vowel: Option<Vowel>,
    pub is_virama: bool,
    pub ayogavaha: Option<Ayogavaha>,
    pub svara: Option<Svara>,
}

By decomposing text into discrete AksharaToken instances:

• Consonant conjuncts (saṁyogākṣara) are preserved across complex scripts (e.g. Grantha ligatures, Malayalam chillu forms, Bengali conjuncts).
• Vedic combining accents are attached directly to the vocalic nucleus of the syllable, guaranteeing that they never become orphan non-spacing marks.

---

2. Vedic Pitch Accents (Svara) Across Scripts

In the Vedic oral tradition, pitch accent (Svara) is phonemic: changing the tone changes the meaning of the word (e.g., Pāṇini Aṣṭādhyāyī 1.2.29–32 and Patañjali’s Mahābhāṣya on the danger of misplacing an accent: śabdo hīnaḥ svarato varṇato vā…).

The Four Primary Vedic Svaras

Svara Name	Musical / Acoustic Pitch	Devanagari Glyph	Unicode Codepoint	Telugu / Kannada Representation
Udātta	High (acute)	Unmarked	—	Unmarked
Anudātta	Low (grave)	Subscript bar _	U+0952	Subscript bar (\u{0952} or script equivalent)
Svarita	Falling circumflex	Superscript vertical stroke |	U+0951	Superscript vertical stroke (\u{0951})
DīrghaSvarita	Long falling circumflex	Double superscript stroke ||	U+1CDA	Double vertical stroke (\u{1CDA})

Cross-Script Svara Preservation

Most Indic script engines strip accents when converting to South Indian scripts because Telugu and Kannada fonts historically used distinct code ranges.

In vyasa-lipi, combining marks U+0951 and U+0952 are canonical Vedic combining marks specified by the Unicode standard to apply across all Indic abugidas:

# Devanagari:
अ॒ग्निमी॑ळे पु॒रोहि॑तम्

# Transliterated to Telugu:
అ॒గ్నిమీ॑ళే పు॒రోహి॑తమ్

# Transliterated to Kannada:
ಅ॒ಗ್ನಿಮೀ॑ಳೆ ಪು॒ರೋಹಿ॑ತಮ್

Notice that both the Anudātta under-bar (\u0952) on a and ro, and the Svarita stroke (\u0951) on mī, remain intact and correctly positioned over the respective Telugu and Kannada aksharas.

---

3. Strict ISO 15919 vs IAST

One of the most persistent errors in digital Sanskrit tools is treating IAST (International Alphabet of Sanskrit Transliteration) and ISO 15919 as synonymous.

Key Divergences

Phonetic Category	Sanskrit Sound	Strict ISO 15919	Standard IAST	Why the Difference Matters
Vocalic Liquid R	ऋ	r̥ (under-ring)	ṛ (under-dot)	Under-ring designates vocalic syllabicity. Under-dot is reserved for retroflex consonants (ṭ, ḍ, ṣ).
Long Vocalic R	ॠ	r̥̄ (ring + macron)	ṝ (dot + macron)
Vocalic Liquid L	ऌ	l̥ (under-ring)	ḷ (under-dot)	Critical: In IAST, ḷ is ambiguous—it is used for both vocalic liquid ऌ and the Vedic retroflex consonant ळ!
Dravidian Short E	ऎ (Telugu/Kannada)	e	N/A	ISO 15919 cleanly distinguishes short e from long ē.
Sanskrit Long E	ए	ē (with macron)	e (bare)	In Classical Sanskrit, e is always long; IAST leaves it bare, which fails when transcribing South Indian texts.
Dravidian Short O	ऒ	o	N/A
Sanskrit Long O	ओ	ō (with macron)	o (bare)
Vedic Lateral Flap	ळ	l̤ (under-diaeresis)	ḷ	ISO 15919 prevents confusion with vocalic ऌ.

vyasa-lipi enforces this distinction strictly:

• If target is iso15919, it emits r̥, r̥̄, l̥, ē, ō, and l̤.
• If target is iast, it emits ṛ, ṝ, ḷ, e, and o.

---

4. Grantha vs Tamil Phonology

Users often ask: Why does vyasa-lipi support Grantha, but not standard Tamil?

The answer lies in the deep history of South Indian epigraphy and phonology.

The Dravidian Consonant Collapse in Tamil

The Classical Tamil script (evolved from Tamil-Brahmi) was engineered strictly for the phonetic structure of Old Tamil, as codified in the Tolkāppiyam.

In Dravidian phonology, there are no distinct phonemic registers for voiced or aspirated stops:

• Sanskrit 4-register stops: k, kh, g, gh
• Tamil phonemic inventory: k (க)

When Sanskrit is transcribed into unaugmented Tamil:

• Govinda becomes Kovinta (கோவிந்த).
• Bhārata becomes Pārata (பாரத).
• Dharma becomes Taruma (தரும).

Why Lossy Mode Violates vyutils Invariants

If vyasa-lipi transliterated Devanagari to standard Tamil:

• धर्म ⟶ தர்ம ⟶ तर्म
• भारत ⟶ பாரத ⟶ पारात

The round-trip fidelity is destroyed because information was lost (gh, g, kh → k).

The Solution: Grantha (𑌗𑍍𑌰𑌨𑍍𑌥)

In the 6th–7th century CE, the Pallava dynasty recognized this exact problem: they needed to write sacred Sanskrit texts in the Tamil-speaking region without losing phonetic distinctions.

They developed Grantha, a sister script to Tamil sharing the same aesthetic ductus, but retaining the complete 33-consonant, 4-stop inventory of Sanskrit.

flowchart TD
    SouthernBrahmi[Southern Ashokan Brahmi] --> PallavaGrantha[Pallava Script]
    PallavaGrantha --> TamilScript[Modern Tamil Script: 1 Stop Register]
    PallavaGrantha --> GranthaScript[Grantha Script: Full Sanskrit Phonology]
    PallavaGrantha --> MalayalamScript[Malayalam Script: Full Sanskrit Phonology]

In vyasa-lipi, Grantha is the canonical script for transcribing Sanskrit in the Tamil cultural sphere (Unicode range 11300..1137F).

---

5. Computational ASCII Schemes

For programmatic pipelines, command-line search, and database indexing where Unicode combining marks can be awkward, vyasa-lipi provides three standard ASCII transliterations:

1. SLP1 (Sanskrit Library Phonetic Basic):
   • 1:1 single-character mapping (every Sanskrit sound maps to exactly one ASCII character).
   • Capital letters denote aspirates and long vowels (A = ā, K = kh, G = gh).
   • Standard in the Cologne Digital Sanskrit Dictionaries.
2. Harvard-Kyoto (HK):
   • Widely used by Western Indologists.
   • Long vowels capitalized (A, I, U). Sibilants: z = ś, S = ṣ.
3. WX:
   • Developed at IIT Kanpur and University of Hyderabad.
   • Specifically engineered for Indian Language to Indian Language Machine Translation (ILMT) algorithms.

---

Next Steps

• Try hands-on transliterations: How-To Guide: Transliterating Sanskrit
• Review CLI arguments and script flags: vyasa-lipi Reference