1. The Phonological-Orthographic Mismatch
The Sinhala writing system belongs to the Brahmic family of South Asian abugidas. Classical Sinhala orthography distinguishes four distinct graphemes:
- දන්තජ න (Dental 'න' - U+0DB1): Historically articulated with the tongue tip against the upper teeth [n].
- මුර්ධජ ණ (Retroflex 'ණ' - U+0DAE): Historically articulated with the tongue curled back against the hard palate [ɳ].
- දන්තජ ල (Dental 'ල' - U+0DBD): Articulated as a dental lateral approximant [l].
- මුර්ධජ ළ (Retroflex 'ළ' - U+0DC5): Historically articulated as a retroflex lateral approximant [ɭ].
Sometime during the transitional Middle Sinhala period (10th–14th centuries CE), the phonemic contrast between dental and retroflex pronunciations was neutralized in colloquial speech. In modern Sri Lankan pronunciation, both න and ණ represent identical [n] sounds, and both ල and ළ represent identical [l] sounds.
2. Classical Etymological Rules and Modern Errors
Because pronunciation offers no auditory guide, writers must rely on historical etymological memory or grammatical rules. Classical treatises like the 13th-century Sidath Sangarawā codified strict sandhi and assimilation constraints:
However, in digital communication, social media, and journalism, confusion abounds:
| Erroneous Digital Form | Correct Etymological Form | Linguistic Etymology / Rationale |
|---|---|---|
| ගුනාංග (gunānga)* | ගුණාංග (guṇānga) | Derived from Sanskrit guṇa (virtue/attribute) requiring retroflex ණ. |
| කල්යානය (kalyānaya)* | කල්යාණය (kalyāṇaya) | Retroflex ණ mandated after palatalized clusters in tatsama loanwords. |
| ප්රතිපල (pratipala)* | ප්රතිඵල / ප්රතිපල (pratiphala) | Aspiration and dental lateral consistency with Pali/Sanskrit phala. |
| ළඟ (langa with dental)* | ළඟ (ḷanga) | Tadbhavic retention of initial retroflex lateral in spatial deictics. |
3. The Computational Disambiguation Architecture
The Sodupath engine integrates a tripartite verification system to detect and rectify Na/La infractions in real-time:
- Etymological Tri-Trie Lexicon: A verified lemma dictionary compiled from the Sinhala Dictionary Office, containing over 75,000 canonical roots mapped to their historical classifications (Tatsama, Tadbhava, Deshya).
- Contextual Morphological Participle Scanner: Differentiating past participles (e.g., දුන් [dun] from root දෙ-) from homophonic noun compounds, ensuring false corrections are not applied to verbal stems.
- Phonetic Levenshtein Distance with Grapheme Clustering: Measuring edit distances strictly within the dental-retroflex permutation subspace (e.g., cost(න, ණ) = 0.25, while cost(න, ක) = 1.0), prioritizing orthographic fixes over semantic changes.