Kurdish Speech is the first research and knowledge enterprise to develop Automatic Speech Recognition (ASR), Speaker Recognition, and Speech Command software for Kurdish through Artificial Intelligence and Signal Processing.
Kurdish Speech is a research and business group working on natural language processing technologies for Kurdish language. Kurdish is a member of the Indo-Iranian branch of Indo-European languages spoken by over 40 million people mainly in Iraq, Turkey, Iran, Syria, Armenia, and Azerbaijan. Despite its diversity of dialects, Kurdish belongs to low-resourced languages in computational linguistics. Kurdish Speech conducts continuous research to develop computational resources, models, and real-world NLP applications.
Synthesizing human-like, natural voice audio from Kurdish texts.
High-precision transcription of spoken Central Kurdish into text.
Accurate orthographic error detection and normalization (AsoSpell).
Kurdish Speech is a knowledge-based enterprise founded and managed by experts in Artificial Intelligence, Computer Engineering, and Computational Linguistics.
We actively build essential resources including text corpora, speech corpora, Kurdish WordNet, phonetic lexicons, and parallel translation datasets.
Over 30 hours of verified Kurdish speech data has been carefully recorded, tagged, and transcribed for speech recognition, phonology, and dialect analysis.
Key artificial intelligence and NLP concepts researched by our engineering team
Enabling computers to interpret text and speech. Covers Tokenization, Lemmatization, POS Tagging, NER, and Chunking.
Clarifying the architectural differences and overlaps between Generative AI, Large Language Models, and Foundation Models.
Techniques including Zero-Shot, Few-Shot, Chain-of-Thought (CoT), Tree of Thoughts (ToT), and Retrieval Augmented Generation (RAG).
Parameter-Efficient Fine-Tuning methods (Prompt-tuning, P-tuning) to improve pretrained LLM performance with minimal compute cost.
Live tools developed and supported by Kurdish Speech for developers and linguists
Real-time Text-to-Speech and Speech-to-Text bot for Central Kurdish available directly on Telegram.
Telegram IntegrationPrecise, fast spelling & punctuation corrector. Identifies errors and suggests unified Kurdish standard forms.
Spell CheckerDecomposes Central Kurdish words into constituent morphemes, prefixes, stems, and suffixes (CKMorph).
MorphologyGenerates inflectional forms of Kurdish words based on linguistic morphological rules.
Rule-based NLPGrapheme-to-Phoneme converter for accurate phonetic transcriptions needed for high-quality speech synthesis.
PhoneticsAutomatic meter and rhythm classification of classical and modern Kurdish poems using machine learning.
Poetry AnalysisStay updated with research publications and latest advancements in generative AI
Common questions regarding speech synthesis, recognition, and editing
Realistic Kurdish Text to Speech. TTS stands for Text-to-Speech. It is a technology that converts written text into spoken words. TTS systems take input text and use synthetic voices to produce spoken audio. These systems are widely used in accessibility features, automated customer service, navigation systems, and audiobook generation.
Automatic Speech Recognition for Central Kurdish. ASR converts spoken language into written text (STT). ASR systems analyze audio input and transcribe it into textual output, making it possible to convert spoken words into a digital format that can be processed, stored, or analyzed by computers.
Precise and Fast for Editing Kurdish Texts. Flags spelling errors and makes suggestions for correction, flags punctuation errors and fixes them at once, detects various spellings of words, and suggests the unified standard form.
Reach out to our research and engineering team
Iran, Tehran, Vanak
(Iran) +98 912 505 3609