1. The Morphological Ambiguity of Terminal 'න්'
In inflectional Indo-Aryan languages, homophony between inflectional morphemes and derivational or root-final codas presents severe hurdles for automated syntactic parsers. In written Sinhala, the virama consonant 'න්' (Hal-Na) occurs in four entirely distinct syntactic categories:
| Syntactic Category | Representative Tokens | Grammatical Role | Expected SVA Behavior |
|---|---|---|---|
| 1. Animate Plural Noun | මිනිසුන්, ළමුන්, සිසුන් | Nominative / Accusative Subject/Object | Demands Plural Verb (කියවූහ, ගත්හ) |
| 2. Verbal Relative Participle | දුන්, හුන්, රැඳුන්, සිටි | Adjectival modifier qualifying following noun | NEUTRAL (Subject governed by Agent) |
| 3. Descriptive Adjective | නොසන්සුන්, සන්සුන්, පිරිපුන්, ගිලන් | Pre-nominal adjective | NEUTRAL (Subject governed by Head Noun) |
| 4. Inanimate Object Noun | අසුන් (seat), රන් (gold), දන් (alms), නිවන් (Nirvana) | Accusative Inanimate Direct Object | NEUTRAL (Verb agrees with Nominative Subject) |
2. The Subject Precedence Architecture
In canonical Sinhala Subject-Object-Verb (SOV) structure, consider the sentence:
ඔහු (He [Singular Masc.]) +
අසුන් (Seat [Inanimate Object]) +
ගත්තේය (Took [Singular Masc. Finite Verb]).
A naive parser examining the token immediately preceding the verb detects අසුන් (ending in -ුන්), erroneously flags it as a plural subject, and generates a spurious error recommending plural verb ගත්තාහුය / ගත්හ.
Sodupath resolves this through Nominative Agent Precedence (NAP):
- The parser identifies the primary animate agent ඔහු via a verified pronouns and nominative noun whitelist.
- Subsequent tokens ending in 'න්' are scrutinized against the
non_plural_n_wordsand inanimate noun index. - If an intervening token is classified as an inanimate object or relative participle, it is disqualified from subject candidacy, guaranteeing that the terminal verb conforms strictly to the primary agent.