fast-langdetect 🚀

Python 3.9-3.12 support only. 🐍

80x faster and 95% accurate language identification with Fasttext 🏎️

This library is a wrapper for the language detection model trained on fasttext by Facebook. For more information, please visit: https://fasttext.cc/docs/en/language-identification.html 📘

This repository is patched from zafercavdar/fasttext-langdetect, adding multi-language segmentation and better packaging support. 🌐

Facilitates more accurate TTS implementation. 🗣️

Need 200M+ memory to use low_memory mode 💾

Installation 💻

pip install fast-langdetect

Usage 🖥️

For more accurate language detection, please use detect(text,low_memory=False) to load the big model.

Model will be downloaded in /tmp/fasttext-langdetect directory when you first use it.

from fast_langdetect import detect_langs

print(detect_langs("Hello, world!"))
# EN

print(detect_langs("Привет, мир!"))
# RU


print(detect_langs("你好，世界！"))
# ZH

Advanced usage 🚀

from fast_langdetect import detect, parse_sentence, detect_multilingual

print(detect("Hello, world!"))
# {'lang': 'en', 'score': 0.1520957201719284}

print(detect_multilingual("Hello, world!你好世界!Привет, мир!"))
# [{'lang': 'ru', 'score': 0.39008623361587524}, {'lang': 'zh', 'score': 0.18235979974269867}, {'lang': 'ja', 'score': 0.08473210036754608}, {'lang': 'sr', 'score': 0.057975586503744125}, {'lang': 'en', 'score': 0.05422825738787651}]

print(parse_sentence("你好世界！Hello, world！Привет, мир！"))
# [{'text': '你好世界！Hello, world！', 'lang': 'ZH', 'length': 18}, {'text': 'Привет, мир！', 'lang': 'UK', 'length': 12}, {'text': '', 'lang': 'EN', 'length': 0}]

Accuracy 🎯

References to the benchmark