---
language:
- ko
license: apache-2.0
tags:
- ner
- token-classification
- korean
- place-search
- bert
library_name: transformers
pipeline_tag: token-classification
---

# SpotSync NER - Korean Place Search NER Model

Korean Named Entity Recognition model used in the **SpotSync** intelligent place search pipeline.
Extracts **Location (LOC)**, **Brand (BRAND)**, **Category (CAT)**, and **Attribute (ATTR)** from natural language queries.

## Labels

| Label | Description | Example |
|-------|-------------|---------|
| `B-LOC` / `I-LOC` | Location/region | Hongdae, Gangnam, Sinchon Station |
| `B-BRAND` / `I-BRAND` | Brand/store name | Starbucks, McDonald's |
| `B-CAT` / `I-CAT` | Business category | Cafe, Restaurant, Gym |
| `B-ATTR` / `I-ATTR` | Attribute/characteristic | Quiet, Good atmosphere, 24 hours |

## Usage

```python
from transformers import AutoTokenizer, AutoModelForTokenClassification
import torch

model_id = "ille255/spotsync-ner"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForTokenClassification.from_pretrained(model_id)

text = "Hongdae nearby cozy cafe"
tokens = text.split()
inputs = tokenizer(tokens, is_split_into_words=True, return_tensors="pt")

with torch.no_grad():
    outputs = model(**inputs)

predictions = torch.argmax(outputs.logits, dim=2)[0]
id2label = model.config.id2label
for token, pred in zip(tokens, predictions[1:-1]):
    print(f"{token}: {id2label[pred.item()]}")
```

## Model Architecture

- **Base**: BERT (BertForTokenClassification)
- **Hidden Size**: 768
- **Attention Heads**: 12
- **Layers**: 12
- **Vocab Size**: 32,000

## Related Links

- ONNX Quantized version: [ille255/spotsync-ner-onnx](https://huggingface.co/ille255/spotsync-ner-onnx)
- GitHub: [SpotSync Project](https://github.com/IlleJiViN/comp_team)