Turkish NLP Engine
Advanced Turkish natural language processing: morphological analysis, entity recognition, sentiment.
4 min read4,500 viewsUpdated 2026-01-30
NovaGuide Assistant
In this article you'll learn about Overview, then explore Features, and finally review How to use.
Overview
The Turkish NLP Engine provides production-grade natural language understanding for Turkish text. It handles the complexities of Turkish morphology — agglutination, vowel harmony, and case suffixes — out of the box.
Turkish is one of the most morphologically complex languages. Our engine was trained on 50M+ Turkish documents and achieves 96.4% accuracy on standard NLU benchmarks.
Features
| Feature | Accuracy | Speed |
|---|---|---|
| Tokenization & Stemming | 98.2% | < 5ms |
| Named Entity Recognition | 96.4% | < 10ms |
| Sentiment Analysis | 94.1% | < 8ms |
| Intent Classification | 95.7% | < 12ms |
| Language Detection | 99.1% | < 2ms |
- Tokenization & Stemming — accurate word boundary detection and root extraction
- Named Entity Recognition — identifies people, places, organizations, dates, and amounts
- Sentiment Analysis — classifies text as positive, negative, or neutral with confidence scores
- Intent Classification — maps user messages to predefined intents
How to use
Install the skill and assign it to any agent that processes Turkish text. The engine runs automatically — no configuration needed.
You can fine-tune the model with your own training data:
{
"training_data": [
{
"text": "Yarın saat 3'te masa ayırtmak istiyorum",
"intent": "reservation.create",
"entities": [
{ "type": "date", "value": "yarın" },
{ "type": "time", "value": "saat 3" }
]
}
]
}
Fine-tuning typically requires 50–100 examples per intent to see significant improvement.
skillnlpturkish
Was this helpful?
More questions? Contact support