README.md

---
title: Manipulative Detector
colorFrom: red
colorTo: red
sdk: streamlit
app_file: src/streamlit_app.py
tags:
- streamlit
pinned: false
short_description: manipulative-detector
license: other
---

Manipulative Language Detector
This is a text classification model based on multilingual BERT (mBERT), used to identify manipulative language in text. This model has been trained on specially labeled data and is suitable for detecting the following types of language features:

---

Information
Model Name: 'LilithHu/mbert-manipulative-detector
base model: 'bert-base-multilingual-cased'
Model size: 279M parameters (Safetensors format)
Training platform: Google Colab + Hugging Face Hub
Training language: Supports Multilingual (Chinese + English)
Inference task type: text-classification
Number of tag categories: 2 (Non-manipulative/manipulative


---

Web Demo

You can directly access the Web UI of this model:
[https://huggingface.co/spaces/LilithHu/manipulative-detector](https://huggingface.co/spaces/LilithHu/manipulative-detector)

---

Public Inference API

This model is open to all, and Hugging Face inference interface can be used without API Key


----
Python：

from transformers import pipeline

pipe = pipeline("text-classification", model="LilithHu/mbert-manipulative-detector")
result = pipe("your text")
print(result)
```

---
cURL：

```bash
curl -X POST https://api-inference.huggingface.co/models/LilithHu/mbert-manipulative-detector \
  -H "Content-Type: application/json" \
  -d '{"inputs": "your text"}'
```

---

Disclaimer:

This model is for research and educational purposes only and should not be used as a basis for decision-making in serious scenarios such as medical care, law, psychology, and criminal investigation.
The output results are only based on the statistical learning outcomes from the training data and may contain biases or misjudgments. Please maintain critical thinking and do not take the results of this model as the final conclusion or basis.
The creator of the model shall not be held responsible for any consequences resulting from the use of this model.