app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

# 加载模型
model_name = "LilithHu/mbert-manipulative-detector"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

# 二分类标签（非操纵性是0，操纵性是1）
labels = ["Non-manipulative / 非操纵性", "Manipulative / 操纵性"]

def classify(text):
    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
    with torch.no_grad():
        outputs = model(**inputs)
        probs = torch.softmax(outputs.logits, dim=1)
        pred = torch.argmax(probs, dim=1).item()
        confidence = probs[0][pred].item()
        return f"🧠 预测 / Prediction: {labels[pred]}\n📊 置信度 / Confidence: {confidence:.2%}"

# Gradio 界面
interface = gr.Interface(
    fn=classify,
    inputs=gr.Textbox(lines=4, placeholder="Enter text in English or Chinese... / 输入中文或英文句子"),
    outputs="text",
    title="🔍 Manipulative Language Detector / 操纵性语言识别器",
    description="🧪 输入英文或中文句子，系统将判断其是否包含操纵性语言。\nEnter a sentence in English or Chinese to detect if it's manipulative.",
    examples=[
        ["If you really cared, you'd do what I say."],
        ["你不爱我就证明给我看！"],
        ["今天的天气真不错。"]
    ]
)

interface.launch()