dv-normalizer 0.1.9

Dhivehi text normalization for TTS frontends

dv-normalize

A Dhivehi text normalizer for TTS frontends. Converts numbers, dates, times, fractions, scores, abbreviations, money, percentages, and other non-Thaana input into spoken-form Dhivehi.

Status: 0.1.9. The v1 rewrite ships under the existing 0.1.x line. The new public API (normalize, Normalizer, NormalizerConfig) is the supported entry point. The legacy 0.1.x classes are still exported but emit a DeprecationWarning and will be removed in a future major release.

Installation

pip install dv-normalize

Quick start

from dv_normalize import normalize

normalize("ވަކި ލާރިން ވެސް 232.23 ލާރި ހޯދައެވެ")
# 'ވަކި ލާރިން ވެސް ދުއިސައްތަ ތިރީސް ދޭއް ޕޮއިންޓް ދޭއް ތިނެއް ލާރި ހޯދައޭ'

normalize("ޑރ. އިބްރާހިމް 14:30 ގައި އައި")
normalize("ކ.އަތޮޅު ވިލިނގިލިން 120 ކިލޯ މީޓަރު")
normalize("ފޭސް2ގެ")

For repeated use, hold onto a Normalizer instance:

from dv_normalize import Normalizer, NormalizerConfig

n = Normalizer(NormalizerConfig(keep_punctuation=False))
n("ހެލޯ، ދުނިޔެ")  # → 'ހެލޯ ދުނިޔެ'

What it handles

Class	Example input	Example output
Cardinal	232	ދުއިސައްތަ ތިރީސް ދޭއް
Comma-grouped	104,880	(single cardinal, not per-digit)
Per-digit	9982711	spelled digit-by-digit (7+ digit identifier)
Decimal	232.23	ދުއިސައްތަ ތިރީސް ދޭއް ޕޮއިންޓް ދޭއް ތިނެއް
Year	2024	ދެހާސް ސައުވީސް
Year range	1982 - 2024	… ން … އަށް
Time	14:30	ސާދަ ގަޑި ތިރީސް
Ordinal	11ވަނަ	adnominal head form
Fraction	1/2	ދެބައިކުޅަ އެއްބައި
Mixed fraction	1 1/2	… އަދި …
Percent	25%	ފަންސަވީސް ޕަސެންޓް
Oblique ref	2024/3	… ޚާއްސަ <denom-ordinal>
Score	3-2, 0-0, 5-0	compact draw / shutout forms
Money	41,800 ރ	ސާޅީސް އެއްހާސް އަށް ސަތޭކަ ރުފިޔާ
Abbreviation	ޑރ.	ޑޮކްޓަރު (real abbreviations only)
Atoll code	ގދ., ހއ.	ގާފު ދާލު, ހާ އަލިފު (per-letter spelling)
Compound abbrev	ސ.ޢ.ވ.	ޞައްލަﷲ ޢަލައިހި ވަސައްލަމް
Calendar marker	2026 މ., 1447 ހ.	… މީލާދީ, … ހިޖުރީ
Sentence ending	ހޯދައެވެ	ހޯދައޭ (113 rules, context-sensitive)

The classifier is priority-ranked, so more specific patterns (calendar markers, multi-letter compound abbreviations, year ranges) shadow the generic ones. Tokens that don't match any rule pass through unchanged.

Configuration

NormalizerConfig(
    dialect="spoken",            # only option for now
    unknown_latin="passthrough", # "passthrough" | "drop" | "spell"
    decimal_separator="auto",    # "auto" | "dot" | "comma"
    time_system="auto",          # "auto" | "12" | "24"
    currency_default="MVR",
    keep_punctuation=True,
    diagnostic=False,
    strict=False,
)

Diagnostic mode

Normalizer.trace(text) returns the classified token list instead of joined text. Useful for debugging which rule fired:

for tok in Normalizer().trace("ޑރ. އިބްރާހިމް 2024ގައި"):
    print(tok.cls, tok.text, tok.spoken, tok.fields)

Legacy API

The original 0.1.x classes (DhivehiNumberConverter, DhivehiTimeConverter, DhivehiYearConverter, DhivehiTextProcessor) are still importable from dv_normalize but emit a DeprecationWarning. They are scheduled for removal in a future major release — migrate to normalize() / Normalizer.

License

MIT — see LICENSE.