Crate shabda 

shabda — Grapheme-to-Phoneme Conversion

shabda (Sanskrit: word / sound) provides text-to-phoneme conversion for vocal synthesis. It bridges the gap between text input and svara’s phoneme sequences — turning “hello world” into synthesizable phoneme events.

Architecture

Input text
    |
    v
Normalizer (lowercase, expand numbers, handle punctuation)
    |
    v
Tokenizer (split into words, detect sentence boundaries)
    |
    v
G2P Engine (dictionary lookup → rule-based fallback)
    |
    v
Prosody Mapper (stress, intonation from punctuation/syntax)
    |
    v
Vec<PhonemeEvent> (ready for svara)

Quick Start

use shabda::prelude::*;

let g2p = G2PEngine::new(Language::English);
let events = g2p.convert("hello world").unwrap();
// events is a Vec<svara::sequence::PhonemeEvent> ready for rendering

Feature Flags

Feature	Default	Description
std	Yes	Standard library. Disable for no_std + alloc
logging	No	Structured logging via tracing-subscriber
json	No	JSON serialization via serde_json
varna	No	Phoneme inventory validation and language detection via varna
full	No	All of the above

Modules

arpabet

ARPABET-to-svara phoneme mapping.

dictionary

Pronunciation dictionary for common/irregular words.

engine

The G2P engine — ties normalization, dictionary, rules, and prosody together.

error

Error types for the shabda crate.

heteronym

Heteronym disambiguation using preceding-word context.

normalize

Text normalization: lowercasing, punctuation handling, number expansion.

prelude

Convenience re-exports for common usage.

prosody

Prosody assignment from text features.

rules

Rule-based grapheme-to-phoneme conversion for English.

ssml

SSML subset parser for speech synthesis markup.

syllable

Syllabification of phoneme sequences.