app.py

import gradio as gr
import numpy as np
import matplotlib.pyplot as plt


# ========== Mock 词频函数（避免 HuggingFace 下载大语料） ==========
def get_freq(expr):
    expr = expr.lower().replace(" ", "_")
    return (abs(hash(expr)) % 5000) + 50


# ========== E 曲线：固着（Entrenchment） ==========
def entrenchment_real(freq_val):
    x = np.linspace(0, 10, 200)
    strength = min(2.0, 0.5 + np.log(freq_val + 1) / 4)
    shift = 5
    y = 1 / (1 + np.exp(-strength * (x - shift)))
    return x, y


# ========== C 曲线：常规化（Conventionalization） ==========
def conventionalization_real(freq_val):
    x = np.linspace(0, 10, 200)
    spread = min(2.0, 0.4 + np.log(freq_val + 1) / 5)
    shift = 4
    y = 1 / (1 + np.exp(-spread * (x - shift)))
    return x, y


# ===================== 单表达分析 =====================
def analyze_ec(expr):
    f = get_freq(expr)

    entrenchment_text = (
        "### Entrenchment（固着）\n"
        f"表达 **'{expr}'** 的模拟语料频次为：**{f} 次**。\n\n"
        "词频越高，固着曲线越陡峭，越容易在个体大脑中形成自动化加工。\n"
    )

    conventionalization_text = (
        "### Conventionalization（常规化）\n"
        "高频表达更容易在社群中传播与扩散，因此常规化曲线会更早上升、扩散更快。\n"
    )

    summary = (
        "### 综合解释\n"
        f"表达 **'{expr}'** 的频率为 **{f}**。高频使用会带来：\n\n"
        "- 更快的固着（E）→ 心理加工速度更快；\n"
        "- 更快的常规化（C）→ 社群接受与扩散更快；\n\n"
        "这反映了 Schmid E&C 模型的核心：语言使用频率驱动固着与常规化共同形成语言系统的稳定性。"
    )

    # --- E 曲线 ---
    x1, y1 = entrenchment_real(f)
    fig_e = plt.figure(figsize=(4.5, 3.2))
    plt.plot(x1, y1, linewidth=2)
    plt.title("Entrenchment Curve")
    plt.xlabel("Usage Frequency")
    plt.ylabel("Cognitive Strength")
    plt.tight_layout()

    # --- C 曲线 ---
    x2, y2 = conventionalization_real(f)
    fig_c = plt.figure(figsize=(4.5, 3.2))
    plt.plot(x2, y2, linewidth=2, color="#ff8c42")
    plt.title("Conventionalization Curve")
    plt.xlabel("Time / Spread")
    plt.ylabel("Community Adoption")
    plt.tight_layout()

    return entrenchment_text + "\n" + conventionalization_text, fig_e, fig_c, summary


# ===================== 双表达对比 =====================
def compare_two(expr1, expr2):
    f1 = get_freq(expr1)
    f2 = get_freq(expr2)

    # --- E 对比 ---
    xA, yA = entrenchment_real(f1)
    xB, yB = entrenchment_real(f2)

    fig_e = plt.figure(figsize=(5, 3.5))
    plt.plot(xA, yA, linewidth=2, label=f"{expr1} (freq={f1})")
    plt.plot(xB, yB, linewidth=2, label=f"{expr2} (freq={f2})")
    plt.title("Entrenchment Comparison")
    plt.legend()
    plt.tight_layout()

    # --- C 对比 ---
    xC1, yC1 = conventionalization_real(f1)
    xC2, yC2 = conventionalization_real(f2)

    fig_c = plt.figure(figsize=(5, 3.5))
    plt.plot(xC1, yC1, linewidth=2, label=f"{expr1} (freq={f1})")
    plt.plot(xC2, yC2, linewidth=2, label=f"{expr2} (freq={f2})")
    plt.title("Conventionalization Comparison")
    plt.legend()
    plt.tight_layout()

    # --- E 差异热力图 ---
    diff = np.abs(yA - yB)[:50].reshape(10, 5)
    fig_h = plt.figure(figsize=(4, 3))
    plt.imshow(diff, cmap="Reds", aspect="auto")
    plt.colorbar(label="Difference")
    plt.title("E Difference Heatmap")
    plt.tight_layout()

    explanation = (
        "### 双表达对比解释\n\n"
        f"两个表达的模拟词频：\n"
        f"- {expr1}: **{f1} 次**\n"
        f"- {expr2}: **{f2} 次**\n\n"
        "#### 1. 固着（E）差异\n"
        "- 高频表达曲线更陡峭 → 更快固着。\n"
        f"- 若 {expr1} 曲线整体高于 {expr2}，说明它在认知加工上更自动化。\n\n"
        "#### 2. 常规化（C）差异\n"
        "- 高频表达扩散更快，曲线更左移。\n"
        f"- 若 {expr1} 的扩散曲线高于 {expr2}，说明它更容易成为社群惯例。\n\n"
        "#### 3. 热力图解释\n"
        "- 深红表示两表达的固着差异显著。\n"
        "- 浅色表示两表达的固着趋势接近。\n\n"
        "#### 4. 综合结论\n"
        "词频显著影响固着与常规化，体现了 E&C 模型中‘个体认知 × 社群传播’的动态互动机制。"
    )

    return "", fig_e, fig_c, fig_h, explanation


# ===================== UI =====================
with gr.Blocks(
    css="""
    .card {
        background:white;
        padding:15px;
        border-radius:10px;
        box-shadow:0 3px 10px rgba(0,0,0,0.08);
        margin-bottom:12px;
    }
    body {
        background: linear-gradient(135deg,#eef2ff,#e0f2fe);
    }
    """
) as demo:

    gr.Markdown("# 🧠 E&C 模型互动平台（最终稳定版）")

    # ========== 单表达 Tab ==========

    with gr.Tab("单表达分析"):
        expr = gr.Textbox(label="输入表达")
        btn1 = gr.Button("分析")
        out_text = gr.Markdown()
        out_e = gr.Plot()
        out_c = gr.Plot()
        out_sum = gr.Markdown()

        btn1.click(
            analyze_ec,
            inputs=[expr],
            outputs=[out_text, out_e, out_c, out_sum]
        )

    # ========== 双表达 Tab ==========

    with gr.Tab("双表达对比分析"):

        with gr.Row():
            with gr.Column(scale=1):
                gr.HTML("<div class='card'>")
                expr1 = gr.Textbox(label="表达 A")
                expr2 = gr.Textbox(label="表达 B")
                btn2 = gr.Button("开始对比")
                gr.HTML("</div>")

            with gr.Column(scale=2):

                with gr.Row():
                    with gr.Column():
                        gr.HTML("<div class='card'>")
                        cmp_e = gr.Plot()
                        gr.HTML("</div>")

                    with gr.Column():
                        gr.HTML("<div class='card'>")
                        cmp_c = gr.Plot()
                        gr.HTML("</div>")

                gr.HTML("<div class='card'>")
                heatmap = gr.Plot()
                gr.HTML("</div>")

                gr.HTML("<div class='card'>")
                cmp_text = gr.Markdown()
                gr.HTML("</div>")

        _dummy = gr.Textbox(visible=False)

        btn2.click(
            compare_two,
            inputs=[expr1, expr2],
            outputs=[_dummy, cmp_e, cmp_c, heatmap, cmp_text]
        )

demo.launch()