Sokhan | Persian NLP Framework

Sokhan is a fast, lightweight, and optimized framework for natural language processing (NLP) in Persian that provides developers and researchers with advanced NLP capabilities with high accuracy and fast performance. With full support for the Persian language, this tool is suitable for personal, research, and commercial projects.

Video > Voice > Image > Last IT Text

Install Windows

pip install Sokhan

Install Linux/Mac

python3 -m pip install Sokhan

or

python3 -m pip install Sokhan

Take Easy, Just Do it :

from sokhan.core.normalize import Sokhan

Sokhan = Sokhan()
text = 'من اولین پیام شما هستم به دنیای سخن خوش آمدید !'

Sokhan(text)

CLI
-------------
'من اولین پیام شما هستم به دنیای سخن خوش آمدید !'

Important

Fast, Optimized, No Free, No Needed GPU And Hard Processing.

Lang :

• Support Persian
• Support Engilish

Feature :

• Support CPU/GPU[No Need Requerment]
• Super Fast Proccessing / Turbo Fast Proccessing
• Config With .json Files
• Base C Lang[For Speed UP]
• API Web

Framework Modules :

• Normalizer
• Informal Normalize
• Summerizer
• Twitter Normalizer
• Instagram Normalizer
• Youtube Normilzer
• Telegram Normilzer
• Whatsapp Normalizer

Sokhan Vs AI In this test, the models are tested on graphics cards of at least H100, while the Sokhan library is being tested on the least hardware resources.

Lang/Module's	Sokhan*	GPTv4	DeepSeek	Gemini	Gama	Grok v3	Shiraz
Count Keyword	0.00000001	No	No	No	No	No	Yes

Sokhan Benchmark's language

Lang/Module's	Sokhan*	NLTK	Fast Text	SpaCy	Regex	Scikit-Learn	Hazm
Support Persian	Yes	No	No	No	No	No	Yes
Support Engilish	Yes	No	No	No	No	No	no
Support Arabic	Cooming...	No	No	No	No	No	no
Support Russian	Cooming...	No	No	No	No	No	no
Support France	Cooming...	No	No	No	No	No	no
Support Italy	Cooming...	No	No	No	No	No	no
Support Spanol	Cooming...	No	No	No	No	No	no

Speed Benhmark:

• Cpu 2 Core | No GPU | Ram 8 Gig | Python 3.10
• 10 Gig Dataset's Comment Ronaldo.

Feature's/Module's	Sokhan*	NLTK	Fast Text	SpaCy	Regex	Scikit-Learn	Hazm
Normalize	0.00000016123	-------	-------	-------	-------	-------	-------
informal normalize	0.00000016123	-------	-------	-------	-------	-------	-------
Summrize	0.00000016123	-------	-------	-------	-------	-------	-------
Clean Text	0.00000016123	-------	-------	-------	-------	-------	-------

Dataset's Benchmark:

Dataset's/Module's	Sokhan*	NLTK	Fast Text	SpaCy	Regex	Scikit-Learn	Hazm
Instagram	Yes	-------	-------	-------	-------	-------	-------
Twitter	Yes	-------	-------	-------	-------	-------	-------
Youtube	Yes	-------	-------	-------	-------	-------	-------
Telegram	Yes	-------	-------	-------	-------	-------	-------
Bale.io	Yes	-------	-------	-------	-------	-------	-------
Eeta	Yes	-------	-------	-------	-------	-------	-------
Tiktok	Yes	-------	-------	-------	-------	-------	-------
Robika	Yes	-------	-------	-------	-------	-------	-------
Hamshahri	Yes	-------	-------	-------	-------	-------	-------
Forbs	Yes	-------	-------	-------	-------	-------	-------

requerment's :

• Python 3.10
• Cython

Suitable for:

• Researcher
• Student's
• Bussiness Productor
• Startup Projects