---
license: cc-by-4.0
language:
- th
base_model:
- airesearch/wangchanberta-base-att-spm-uncased
pipeline_tag: token-classification
---
library_name: transformers
tags: [ner, thai, food, review, token-classification]
---

# Model Card for wttw/modchelin_thainer-base-model

This model performs Named Entity Recognition (NER) on Thai-language food reviews. It is designed to extract domain-specific aspects such as dish names, ingredients, restaurant service, and sentiment-related phrases from customer-written content.

## Model Details

### Model Description

This is the model card of a 🤗 Transformers model that has been pushed to the Hugging Face Hub.

- **Developed by:** Vitawat Kitipatthavorn  
- **Finetuned from model:** `airesearch/wangchanberta-base-att-spm-uncased`  
- **Model type:** Token Classification (NER)  
- **Language(s) (NLP):** Thai  
- **License:** cc-by-sa-4.0  
- **Shared by:** wttw  
- **Model ID:** `wttw/modchelin_thainer-base-model`

## Uses

### Direct Use

This model is designed for extracting domain-specific entities from Thai-language food reviews. It identifies and classifies named entities related to:

- Food/menu items  
- Taste  
- Service  
- Ambiance  
- Price and value  
- Other aspects relevant to customer dining experiences

**Example:**

- **Input:** `"ต้มยำกุ้งอร่อยมาก แต่บริการช้า"`  
- **Output:**  
  - `ต้มยำกุ้ง: FOOD`  
  - `บริการ: SERVICE`

The model is suitable for NLP pipelines aimed at analyzing restaurant reviews, powering sentiment dashboards, or supporting aspect-based sentiment analysis (ABSA).

### Downstream Use

The model can be integrated into:

- Thai ABSA pipelines  
- Restaurant feedback summarization systems  
- Chatbots or moderation tools for food delivery and review platforms

### Out-of-Scope Use

The model is not designed for:

- Non-food-related documents (e.g., legal, clinical, political)  
- General-purpose Thai NER tasks  
- Use cases requiring high confidence on ambiguous or out-of-domain text

## Bias, Risks, and Limitations

The model is trained specifically on food review content and may:

- Struggle with informal slang or regional dialects  
- Over-predict `FOOD` entities in unrelated contexts  
- Misclassify ambiguous phrases without surrounding context

### Recommendations

Users should:

- Avoid applying this model outside food-related domains  
- Fine-tune further if working with reviews in specific dialects or contexts  
- Evaluate on a sample of target data before production use  
- Consider setting confidence thresholds before using predictions downstream

## How to Get Started with the Model

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

model_name = "wttw/modchelin_thainer-base-model"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForTokenClassification.from_pretrained(model_name)

ner_pipeline = pipeline("ner", model=model, tokenizer=tokenizer, aggregation_strategy="simple")

example = "ต้มยำกุ้งอร่อยมาก แต่บริการช้า"
entities = ner_pipeline(example)

print(entities)