Metadata-Version: 2.4
Name: lectura-stt-formules
Version: 0.1.0
Summary: STT dedie formules — CTC semantique pour nombres, dates, heures, sigles (generateur de corpus)
Author-email: Max Carriere <contact@lec-tu-ra.com>
License: AGPL-3.0-or-later
Project-URL: Homepage, https://www.lec-tu-ra.com/solutions/outils/modules/
Project-URL: Repository, https://github.com/maxcarriere/lectura-modules/tree/main/STT-Formules
Project-URL: Issues, https://github.com/maxcarriere/lectura-modules/issues
Keywords: stt,asr,ctc,french,formulas,numbers,dates
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: Intended Audience :: Science/Research
Classifier: License :: OSI Approved :: GNU Affero General Public License v3 or later (AGPLv3+)
Classifier: Natural Language :: French
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Classifier: Topic :: Multimedia :: Sound/Audio :: Speech
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENCE.txt
Requires-Dist: lectura-formules>=3.7
Provides-Extra: corpus
Requires-Dist: lectura-tts-multispeaker>=1.3; extra == "corpus"
Requires-Dist: numpy>=1.24; extra == "corpus"
Requires-Dist: soundfile>=0.12; extra == "corpus"
Provides-Extra: training
Requires-Dist: torch>=2.0; extra == "training"
Requires-Dist: torchaudio>=2.0; extra == "training"
Requires-Dist: soundfile>=0.12; extra == "training"
Requires-Dist: numpy>=1.24; extra == "training"
Provides-Extra: export
Requires-Dist: torch>=2.0; extra == "export"
Requires-Dist: onnx>=1.14; extra == "export"
Requires-Dist: onnxruntime>=1.16; extra == "export"
Provides-Extra: inference
Requires-Dist: onnxruntime>=1.16; extra == "inference"
Requires-Dist: numpy>=1.24; extra == "inference"
Requires-Dist: soundfile>=0.12; extra == "inference"
Dynamic: license-file

# lectura-stt-formules

STT dedie formules — modele CTC autonome avec vocabulaire semantique
(~87 tokens : nombres atomiques, mois, lettres, marqueurs) au lieu de
phonemes IPA.

## Phase 1 — Generateur de corpus

Ce module fournit :

- Un vocabulaire de 87 tokens semantiques (`_vocab.py`)
- Un tokenizer events → token sequence (`_tokenizer.py`)
- Un generateur de corpus synthetique (`scripts/generate_corpus.py`)

## Installation

```bash
pip install lectura-stt-formules

# Pour la generation de corpus (necessite TTS)
pip install lectura-stt-formules[corpus]
```

## Utilisation

### Vocabulaire et tokenizer

```python
from lectura_stt_formules import VOCAB, events_to_token_sequence, token_ids_to_names
from lectura_formules import lire_nombre

result = lire_nombre("42")
tokens = events_to_token_sequence(result)
print(tokens)           # [22, 4]
print(token_ids_to_names(tokens))  # ['QUARANTE', 'DEUX']
```

### Generation de corpus

```bash
python scripts/generate_corpus.py \
    --output-dir /data/voix_ssd/formula_corpus/ \
    --n-base 16000 \
    --n-augmentations 3 \
    --seed 42 \
    --num-workers 4
```

## Licence

AGPL-3.0-or-later — voir [LICENCE.txt](LICENCE.txt)
