Pengzhendong Text Normalization Save

Text Normalization & Inverse Text Normalization

0. Brief Introduction

- **Must Read Doc** (In Chinese): https://mp.weixin.qq.com/s/q_11lck78qcjylHCi6wVsQ

WeTextProcessing: Production First & Production Ready Text Processing Toolkit

0.1 Text Normalization

0.2 Inverse Text Normalization

1. How To Use

1.1 Quick Start:

# install
pip install WeTextProcessing

Command-usage:

wetn --text "2.5平方电线"
weitn --text "二点五平方电线"

Python usage:

# tn usage
>>> from tn.chinese.normalizer import Normalizer
>>> normalizer = Normalizer()
>>> normalizer.normalize("2.5平方电线")
# itn usage
>>> from itn.chinese.inverse_normalizer import InverseNormalizer
>>> invnormalizer = InverseNormalizer()
>>> invnormalizer.normalize("二点五平方电线")

1.2 Advanced Usage:

DIY your own rules && Deploy WeTextProcessing with cpp runtime !!

For users who want modifications and adapt tn/itn rules to fix badcase, please try:

git clone https://github.com/wenet-e2e/WeTextProcessing.git
cd WeTextProcessing
pip install -r requirements.txt
pre-commit install # for clean and tidy code
# `overwrite_cache` will rebuild all rules according to
#   your modifications on tn/chinese/rules/xx.py (itn/chinese/rules/xx.py).
#   After rebuild, you can find new far files at `$PWD/tn` and `$PWD/itn`.
python -m tn --text "2.5平方电线" --overwrite_cache
python -m itn --text "二点五平方电线" --overwrite_cache

Once you successfully rebuild your rules, you can deploy them either with your installed pypi packages:

# tn usage
>>> from tn.chinese.normalizer import Normalizer
>>> normalizer = Normalizer(cache_dir="PATH_TO_GIT_CLONED_WETEXTPROCESSING/tn")
>>> normalizer.normalize("2.5平方电线")
# itn usage
>>> from itn.chinese.inverse_normalizer import InverseNormalizer
>>> invnormalizer = InverseNormalizer(cache_dir="PATH_TO_GIT_CLONED_WETEXTPROCESSING/itn")
>>> invnormalizer.normalize("二点五平方电线")

Or with cpp runtime:

cmake -B build -S runtime -DCMAKE_BUILD_TYPE=Release
cmake --build build
# tn usage
cache_dir=PATH_TO_GIT_CLONED_WETEXTPROCESSING/tn
./build/processor_main --tagger $cache_dir/zh_tn_tagger.fst --verbalizer $cache_dir/zh_tn_verbalizer.fst --text "2.5平方电线"
# itn usage
cache_dir=PATH_TO_GIT_CLONED_WETEXTPROCESSING/itn
./build/processor_main --tagger $cache_dir/zh_itn_tagger.fst --verbalizer $cache_dir/zh_itn_verbalizer.fst --text "二点五平方电线"

2. TN Pipeline

Please refer to TN.README

3. ITN Pipeline

Please refer to ITN.README

Discussion & Communication

For Chinese users, you can aslo scan the QR code on the left to follow our offical account of WeNet. We created a WeChat group for better discussion and quicker response. Please scan the personal QR code on the right, and the guy is responsible for inviting you to the chat group.

Or you can directly discuss on Github Issues.

Acknowledge

1. Thank the authors of foundational libraries like OpenFst & Pynini.
2. Thank NeMo team & NeMo open-source community.
3. Thank Zhenxiang Ma, Jiayu Du, and SpeechColab organization.
4. Referred Pynini for reading the FAR, and printing the shortest path of a lattice in the C++ runtime.
5. Referred TN of NeMo for the data to build the tagger graph.
6. Referred ITN of chinese_text_normalization for the data to build the tagger graph.