TUNiBrain–Korean Morphological Analyzer

Hey Product Hunters 👋

If you’ve ever built NLP systems for Korean, you already know the ugly truth:

**tokenization alone is not understanding.**

Most analyzers can split text — but they can’t tell whether

**“눈” means an eye or snow**,

**“배” is a fruit, a ship, or a stomach**,

or whether a word is a **person, place, or brand**.

So we built **TUNiBrain**.

TUNiBrain is not just a tokenizer — it’s a **meaning-aware Korean language engine**.

Under the hood, it combines:

- **CRF-based MeCab parsing**
- **Bi-LSTM proper-noun detection**
- **LESK + TF-IDF word-sense disambiguation**
- **A semantic dictionary (TUNiBank)**

This allows it to produce output like:

> “나는 에배레스트에서 배를 먹고 내려왔다”
→ mistyped Everest(에배레스트) is detected as a proper noun
→ “배” is disambiguated as **pear**, not **stomach**
→ each token is assigned **sense ID, definition, origin, and English equivalent**

TUNiBrain

Compared to MeCab-ko or Khaiii, TUNiBrain was built to win where real products fail:

**proper nouns, ambiguity, and meaning**
TUNiBrain

It delivers:

- 🧠 **Word-sense disambiguation (WSD)**
- 🏷 **Proper noun recognition**
- 🔢 **Number detection**
- 📚 **Definitions, origins, translations, semantic & topic tags**
- 🔧 **Retrainable models & customizable dictionaries** TUNiBrain

### 🔓 Open Source, with a Clear Boundary

TUNiBrain is **open source for research and non-commercial use**.

If you’d like to use it in a **commercial product or service**, please contact us — we offer enterprise licenses and support.

• **Attribution required**: Please cite TUNiB Inc. when using this software in academic papers or research publications.

**Commercial inquiries:👉 📩tunibridge@tunib.ai

We already use TUNiBrain inside **TUNiB’s AI Guardrails, chatbots, and enterprise NLP pipelines**, and we’re excited to see what the community builds with it.

If you’re working with Korean text — try it, break it, and tell us what you need.

Let’s make Korean NLP truly intelligent 🧠🚀

— Team TUNiB