Why Eke?
The problem with Kannada romanisation
Kannada has no single accepted romanisation. Writers on the internet borrow whatever feels natural — some use IAST (ā, ṭ, ḷ), others use Hindi-inflected transliterations, and most use something ad hoc. The result: the same word appears dozens of ways, no scheme is searchable, and learners cannot reliably convert between script and roman.
Meanwhile, the two rigorous scholarly systems have real drawbacks:
- IAST — requires diacritic characters (ā, ī, ṭ, ḍ, ṇ, ḷ) that can’t be typed on a standard keyboard and don’t sort or search well in plain text
- Harvard-Kyoto (HK) — designed for Sanskrit, not Kannada; represents anusvara as standalone
M(which is phonologically wrong for Kannada), conflates ಶ and ಷ, and has no natural way to represent Dravidian vowel distinctions
Despite these drawbacks, ASCII-based romanisation has a significant advantage that the native scripts don’t: cross-language readability across Dravidian languages. A reader familiar with HK or Eke can follow Kannada, Tamil, Telugu, and Malayalam texts in romanisation with relatively little additional learning — the same consonant and vowel symbols appear in all four systems. Speakers of non-Dravidian languages who are comfortable with the English alphabet can also adjust to reading Dravidian languages in Eke far more quickly than they could learn a new script.
DNS Bhat’s influence: Ellara Kannada
Dr. D.N. Shankara Bhat’s Ellara Kannada reform argues that modern written Kannada should reflect how the language is actually spoken, not how Sanskrit grammar was once overlaid on it. Native Kannada has no mahaprāṇa (aspirated consonants) — bh, dh, kh etc. are borrowings from Sanskrit that do not appear in Dravidian root words. Eke(ek) — the coining mode — follows this: new native words avoid aspirates.
Hosabaraha orthography
Hosabaraha (Vishwas’s Kannada input system) uses a consistent phone-key mapping for Kannada characters. Its orthographic conventions are the same as Ellara Kannada — Hosabaraha is the Kannada-script counterpart of Eke(ek). Eke formalises this mapping as a romanisation that:
- Works with plain ASCII on any keyboard
- Is completely reversible — every roman string maps back to exactly one Kannada string
Symbol complexity: Classical → Ellara → Eke
There are two ways to measure the burden of a script: the number of distinct symbols a learner must memorise, and the size of the permuted symbol space that must be recognised in context.
Actual symbols (what you must learn)
| Component | Classical | Ellara Kannada | Eke |
|---|---|---|---|
| Main consonants | 35 | 21 (mahāprāṇas and rare letters merged or dropped) | — |
| Ottakṣara (conjunct) consonants | 35 | 21 | — (sequences of letters) |
| Independent vowels | 16 | 10 | — |
| Vowel diacritics (mātrās) | 16 | 10 | — |
| Modifiers | 5 (virama ್, nuktā ಼, anusvara ಂ, visarga ಃ, jihvamūlīya ೱ) | 3 (virama, nuktā, anusvara; drops ಃ and ೱ) | — |
| Total | ~107 | ~65 | 31 / 41 / 46 |
Classical Kannada has 16 vowels (independent form / vowel diacritic):
| IAST | Independent | Diacritic | Retained in Ellara? |
|---|---|---|---|
| a | ಅ | (inherent) | ✓ |
| ā | ಆ | ಾ | ✓ |
| i | ಇ | ಿ | ✓ |
| ī | ಈ | ೀ | ✓ |
| u | ಉ | ು | ✓ |
| ū | ಊ | ೂ | ✓ |
| ṛ | ಋ | ೃ | ✗ — Sanskrit-only vocalic syllabic |
| ṝ | ೠ | ೄ | ✗ — Sanskrit-only, long form |
| ḷ | ಌ | ೢ | ✗ — Sanskrit-only vocalic syllabic |
| ḹ | ೡ | ೣ | ✗ — Sanskrit-only, long form |
| e | ಎ | ೆ | ✓ |
| ē | ಏ | ೇ | ✓ |
| ai | ಐ | ೈ | ✗ — becomes vowel+glide sequence ay |
| o | ಒ | ೊ | ✓ |
| ō | ಓ | ೋ | ✓ |
| au | ಔ | ೌ | ✗ — becomes vowel+glide sequence av |
Ellara Kannada drops 6 of these — the four Sanskrit vocalic-syllabic sounds (ṛ ṝ ḷ ḹ) that do not occur in native Kannada speech, and the two diphthongs ai ಐ and au ಔ which become the vowel+glide sequences ay and av. This leaves 10 vowels: a ಅ, ā ಆ, i ಇ, ī ಈ, u ಉ, ū ಊ, e ಎ, ē ಏ, o ಒ, ō ಓ.
Nuktā (಼) is retained in both Classical and Ellara Kannada for representing sounds from contact languages — fa, za, odd (English), ḵ (Arabic/Urdu). It is also needed for dialect phonemes: Havyaka’s unrounded-u vowel, Toda’s retroflex laterals. In Eke, two additional modifier symbols handle these: : (nuktā diacritic, for sounds like Toda retroflex laterals) and ^ (rounding/unrounding marker, for Havyaka’s unrounded-u: u^). These two bring the full dialect-inclusive count to 46.
Permuted symbol space (what you must recognise)
The ottakṣara system multiplies consonants combinatorially — any consonant can stack with any other. Combined with vowel forms, the recognisable glyph space is:
| Script system | Formula | Space |
|---|---|---|
| Classical Kannada | 35 × 35 × 21 (consonants × conjuncts × (16 diacritics + 5 modifiers)) | ≈ 25,725 |
| Ellara Kannada | 21 × 21 × 13 (21 consonants × conjuncts × (10 diacritics + 3 modifiers)) | ≈ 5,733 (-77%) |
| Eke | no conjuncts; 31 × 1 letter-forms | 31 / 41 / 46 |
Ellara Kannada achieves a 77% reduction with no loss of fidelity for spoken modern Kannada. Sanskrit and Vedic texts require the full classical set.
Eke eliminates the combinatorial explosion entirely: by using a plain alphabet, a consonant sequence is just letters side by side — no stacked glyph, no recognisable conjunct form. The three tiers of coverage:
| Coverage | Symbols |
|---|---|
| Eke(ek) — everyday spoken Kannada (Ellara Kannada equivalent) | 31 |
| Formal Kannada — including historical and rare sounds (full Ellara Kannada) | 41 |
| Extended — adds English f, w, z | 44 |
Dialect-inclusive — adds : (nuktā) and ^ (rounding) for Havyaka, Toda | 46 |
Harvard-Kyoto follows the same alphabet-over-abugida philosophy; Eke makes small modifications to better reflect Kannada phonology.
The two-mode design
A single romanisation can’t serve both archivists and lexicographers. Eke has two modes:
| Mode | For | Aspirates | ಷ |
|---|---|---|---|
| Eke | Romanising existing text | Preserved — bh, Sh, etc. | Sh (distinct from S=ಶ) |
| Eke(ek) | Coining new native words | Dropped | Merged with S |
This lets the same notation system serve both careful transcription of DNS Bhat’s books (which contain Sanskrit loanwords) and the creation of new Dravidian-root vocabulary.
Key design choices
Case = contrast, not emphasis. Uppercase always signals a phonological distinction:
T D N L= retroflexes (vst d n l= dentals)S= palatal ಶ (vss= alveolar ಸ);Sh= retroflex ಷA I U E O= long vowels (vsa i u e o= short)H= visarga ಃ (vsh= ಹ)
Anusvara as assimilation, never M. In Kannada, anusvara ಂ before a plosive is phonologically the nasal of that plosive’s place of articulation — nk, ng, nc, nT, nt, mp, mb. Writing M (as in IAST) is phonologically opaque and creates ambiguity with the labial nasal m.
Diphthongs as sequences. ಐ = ಅ + ಯ್ → ay; ಔ = ಅ + ವ್ → av. These reflect the actual phonological composition in Kannada, not the Sanskrit-influenced ai/au.
This choice is not arbitrary — Kannada has a full productive series of vowel+glide sequences with both ಯ್ and ವ್, across all vowel lengths:
| With ಯ್ | With ವ್ |
|---|---|
ay (ಅಯ್), Ay (ಆಯ್) | av (ಅವ್), Av (ಆವ್) |
iy (ಇಯ್), Iy (ಈಯ್) | iv (ಇವ್), Iv (ಈವ್) |
uy (ಉಯ್), Uy (ಊಯ್) | uv (ಉವ್), Uv (ಊವ್) |
ey (ಎಯ್), Ey (ಏಯ್) | ev (ಎವ್), Ev (ಏವ್) |
oy (ಒಯ್), Oy (ಓಯ್) | ov (ಒವ್), Ov (ಓವ್) |
Sanskrit treats ಐ and ಔ as atomic vowels requiring dedicated symbols (ai, au). Eke rejects this: since iy, Iy, uy, Oy and the rest are all written as plain vowel+glide sequences, ay is simply the a-row entry in the same regular pattern. Writing ai/au would be the only irregularity in the vowel table — a Sanskrit import with no phonological justification in Kannada.
x / q for vocalic syllabics. Sanskrit has four vocalic syllabic sounds that do not occur in native Kannada: vocalic ṛ (ಋ/ೃ), its long form ṝ (ೠ/ೄ), vocalic ḷ (ಌ/ೢ), and its long form ḹ (ೡ/ೣ). None of the four appear in Dravidian roots. Eke maps them to the two ASCII letters that have no other use in Kannada phonology:
| Symbol | Eke | Notes |
|---|---|---|
| ಋ / ೃ | x | short vocalic ṛ — common in Sanskrit loanwords: samskxta, kxShNa |
| ೠ / ೄ | X | long vocalic ṝ — extremely rare |
| ಌ / ೢ | q | short vocalic ḷ — rare; Sanskrit only |
| ೡ / ೣ | Q | long vocalic ḹ — extremely rare |
Fun fact: the x choice echoes history — 𑀋 (U+1000B) is the Asokan Brahmi character for vocalic ṛ, and its visual shape is strikingly close to a lowercase x.
Full specification: Eke.md (153KB, authoritative).