Validator Workflow
Detailed explanation of validation logic and coverage checks.
Sample Test Word List
Curated validation words for diphone coverage and TTS rebuild testing
Validation
Sample Test Word List
This page documents the curated validation word list used to test diphone coverage, safe filename mapping, validator logic, and live TTS playback during system rebuilds.
Purpose of this list.
A small but carefully selected set of words can reveal most system-level problems
in a diphone-based speech synthesizer. These words test boundary diphones,
common consonant–vowel transitions, clusters, and rare phonetic environments.
1. Why a Test Word List Is Important
Running the validator on thousands of words is useful, but debugging becomes difficult. A curated list of about 30–50 representative words makes it easier to isolate rule mismatches and missing diphone files.
Good validation words should test:
- word-initial diphones
- word-final diphones
- common CV transitions
- VC transitions
- nasal environments
- fricatives and affricates
- clusters and learned forms
2. Core Validation Word List
| # | BPM Word | Purpose |
|---|---|---|
| 1 | দিশা | tests d-i and i-ʃ transitions |
| 2 | কথা | k-ɔ and th-a transitions |
| 3 | কাম | k-a and a-m |
| 4 | মান | m-a and a-n |
| 5 | দিন | d-i and i-n |
| 6 | পান | p-a and a-n |
| 7 | গান | g-a and a-n |
| 8 | ভাল | bh-a and a-l |
| 9 | তর | t-ɔ and ɔ-r |
| 10 | ধর | dh-ɔ and ɔ-r |
| 11 | শর | ʃ onset coverage |
| 12 | সার | s onset coverage |
| 13 | হর | h onset coverage |
| 14 | লাল | l onset and coda |
| 15 | রাম | r onset coverage |
| 16 | নাম | nasal onset |
| 17 | মালা | m-a-l transitions |
| 18 | বন | b-ɔ-n sequence |
| 19 | চাল | c onset coverage |
| 20 | ঝাল | jh onset coverage |
| 21 | জল | j onset coverage |
| 22 | টাল | retroflex onset |
| 23 | ডাল | retroflex voiced onset |
| 24 | তাল | dental onset |
| 25 | দাল | dental voiced onset |
| 26 | কাজ | z/j coda environment |
| 27 | গাছ | cʰ/ʧ coda test |
| 28 | রাত | t final diphone |
| 29 | ঘর | gh onset test |
| 30 | নদী | n-ɔ-d-i chain |
| 31 | শিশু | ʃ-i repetition |
| 32 | পথ | p-ɔ-th sequence |
| 33 | ভাষা | bh-a-ʃ-a |
| 34 | দেশ | d-e-ʃ |
| 35 | কাল | k-a-l |
| 36 | গরু | g-ɔ-r-u |
| 37 | চাকা | c-a-k-a chain |
| 38 | নগর | n-ɔ-g-ɔ-r chain |
| 39 | পদ | p-ɔ-d coda |
| 40 | বনজ | cluster-like transitions |
| 41 | কৃষি | cluster environment |
| 42 | প্রকাশ | cluster onset test |
| 43 | ত্রাণ | complex cluster |
| 44 | গ্রাম | cluster onset |
| 45 | শ্রদ্ধা | learned cluster |
| 46 | স্বাধীন | sv cluster test |
| 47 | জ্ঞাত | learned consonant sequence |
| 48 | অন্তর | nasal cluster environment |
| 49 | সংগীত | ŋ environment |
| 50 | বাংলা | ŋ-g transition |
3. How This List Is Used
Run validator ↓ Check diphone sequence ↓ Check safe filenames ↓ Check diphone WAV files ↓ Measure coverage
If this list passes validation, the core TTS system is usually stable enough for wider testing.
4. Recommended Workflow
- run validator on this list
- note missing diphones
- rebuild or segment missing audio
- rerun validator
- confirm stable coverage
5. Related Archive Pages
Rebuild Checklist
Operational guide for rebuilding the diphone inventory safely.
Diphone Inventory
Overview of the full diphone inventory used by the TTS system.
Test list note.
This list should evolve as the dictionary grows and new phonetic environments
are discovered. It should remain small enough to run quickly but broad enough
to detect rule mismatches during rebuilds.