video_decode/examples/static/visualize/tab3.py

#!/usr/bin/env python3
"""
Tab3内容生成器 - 脚本点（元素列表）
"""
import html as html_module
from typing import Dict, Any, List


def calculate_intent_support_count(element: Dict[str, Any]) -> int:
    """
    计算元素的意图支撑数量

    统计元素支撑的意图点总数
    支撑的意图点越多，说明该元素与意图的关联越强

    Args:
        element: 元素数据

    Returns:
        支撑的意图点总数
    """
    # 区分实质元素和形式元素的统计方式
    dimension = element.get('维度') or {}
    if isinstance(dimension, dict) and dimension.get('一级') == '形式':
        # 形式元素：基于多维度评分，使用最高相似度作为"意图支撑强度"的代表
        # 兼容两种结构：
        # 1）新版：{"名称": "...", "相似度结果": [{"点","语义相似度","文本相似度",...}, ...]}
        # 2）旧版：直接列表 [{"点","语义相似度","文本相似度",...}, ...]
        multi_scores = element.get('多维度评分') or {}
        best_score = 0.0

        for point_type in ['灵感点', '目的点', '关键点']:
            for item in multi_scores.get(point_type, []) or []:
                if not isinstance(item, dict):
                    continue

                similarity_results = item.get('相似度结果')

                # 新结构：在相似度结果列表里取最大值
                if similarity_results:
                    for sim in similarity_results or []:
                        if not isinstance(sim, dict):
                            continue
                        semantic = float(sim.get('语义相似度', 0) or 0)
                        text_sim = float(sim.get('文本相似度', 0) or 0)
                        best_score = max(best_score, semantic, text_sim)
                else:
                    # 旧结构：当前item本身就带语义/文本相似度
                    semantic = float(item.get('语义相似度', 0) or 0)
                    text_sim = float(item.get('文本相似度', 0) or 0)
                    best_score = max(best_score, semantic, text_sim)

        return best_score
    else:
        # 实质元素：按意图支撑的数量统计
        intent_support = element.get('意图支撑') or {}

        total_support_count = 0
        for point_type in ['灵感点', '目的点', '关键点']:
            if point_type in intent_support and intent_support[point_type]:
                total_support_count += len(intent_support[point_type])

        return total_support_count


def determine_dominant_factor(element: Dict[str, Any], all_elements: List[Dict[str, Any]]) -> str:
    """
    判断元素排序的主导因素

    排序规则：覆盖率 > 频次 > 意图支撑数
    主导因素判断：哪个指标在当前元素中相对最显著

    Args:
        element: 当前元素
        all_elements: 同组所有元素

    Returns:
        主导因素: 'coverage' | 'frequency' | 'intent_support'
    """
    if not all_elements:
        return 'coverage'

    # 获取当前元素的指标
    commonality = element.get('共性分析') or {}
    coverage = commonality.get('段落覆盖率', 0.0)
    frequency = commonality.get('出现频次', 0)
    intent_count = calculate_intent_support_count(element)

    # 将所有指标归一化到同一量级，然后比较
    # 覆盖率已经是0-1范围
    # 频次归一化：假设最大频次为10
    normalized_frequency = min(frequency / 10.0, 1.0)
    # 意图支撑数归一化：假设最大支撑数为10
    normalized_intent = min(intent_count / 10.0, 1.0)

    # 比较归一化后的值，取最大的作为主导因素
    scores = {
        'coverage': coverage,
        'frequency': normalized_frequency,
        'intent_support': normalized_intent
    }
    return max(scores, key=scores.get)


def sort_elements_by_coverage_and_frequency(elements: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
    """
    按照覆盖率、频次和意图支撑数对元素排序

    排序规则：
    1. 第一优先级：共性（段落覆盖率）- 倒序
    2. 第二优先级：共性（出现频次）- 倒序
    3. 第三优先级：意图支撑数 - 倒序

    Args:
        elements: 元素列表

    Returns:
        排序后的元素列表
    """
    def get_sort_key(elem):
        # 获取共性分析，防止为None
        commonality = elem.get('共性分析') or {}
        # 获取段落覆盖率
        coverage = commonality.get('段落覆盖率', 0.0)
        # 获取出现频次
        frequency = commonality.get('出现频次', 0)
        # 计算意图支撑数
        intent_count = calculate_intent_support_count(elem)
        # 返回排序键（负数用于倒序）
        return (-coverage, -frequency, -intent_count)

    return sorted(elements, key=get_sort_key)


def get_element_category(element: Dict[str, Any]) -> str:
    """
    获取元素的分类名称（支持新旧两种数据结构）

    Args:
        element: 元素数据

    Returns:
        分类名称字符串
    """
    category = element.get('分类', '未分类')
    if isinstance(category, dict):
        # 新结构：分类是对象，包含一级分类和二级分类
        level1 = category.get('一级分类', '')
        level2 = category.get('二级分类', '')
        if level1 and level2:
            return f"{level1} - {level2}"
        elif level1:
            return level1
        else:
            return '未分类'
    else:
        # 旧结构：分类是字符串或列表
        if isinstance(category, list):
            return ' - '.join(category) if category else '未分类'
        return category if category else '未分类'


def group_elements_by_hierarchical_category(elements: List[Dict[str, Any]]) -> Dict[str, Any]:
    """
    按树形分类结构组织元素（一级分类 → 二级分类 → 元素）

    优化规则：同一个父节点下的所有子节点采用统一的分类格式展示
    - 如果一级分类下既有元素又有分类，将元素归入"未分类"二级分类

    Args:
        elements: 元素列表

    Returns:
        树形分类结构字典
    """
    # 1. 按一级分类和二级分类分组
    level1_groups = {}

    for elem in elements:
        category_data = elem.get('分类', {})

        if isinstance(category_data, dict):
            level1 = category_data.get('一级分类', '未分类')
            level2 = category_data.get('二级分类', '')
        elif isinstance(category_data, list):
            # 列表格式：第一个元素作为一级分类，第二个作为二级分类
            level1 = category_data[0] if len(category_data) > 0 else '未分类'
            level2 = category_data[1] if len(category_data) > 1 else ''
        else:
            # 旧结构：分类是字符串
            level1 = str(category_data) if category_data else '未分类'
            level2 = ''

        # 初始化一级分类
        if level1 not in level1_groups:
            level1_groups[level1] = {
                'elements': [],
                'level2_groups': {}
            }

        # 如果有二级分类，放入二级分类组；否则放入一级分类的直接元素列表（临时）
        if level2:
            if level2 not in level1_groups[level1]['level2_groups']:
                level1_groups[level1]['level2_groups'][level2] = []
            level1_groups[level1]['level2_groups'][level2].append(elem)
        else:
            level1_groups[level1]['elements'].append(elem)

    # 1.5 优化：仅当一级分类下既有直接元素又有二级分类时，才将直接元素移到"未分类"二级分类中
    # 如果一级分类下只有直接元素，没有二级分类，则保持原样（不需要"未分类"概念）
    for level1_name, level1_data in level1_groups.items():
        if level1_data['elements'] and level1_data['level2_groups']:
            # 将直接元素移到"未分类"分类
            if '未分类' not in level1_data['level2_groups']:
                level1_data['level2_groups']['未分类'] = []
            level1_data['level2_groups']['未分类'].extend(level1_data['elements'])
            level1_data['elements'] = []
        # 如果只有直接元素，没有二级分类，则保持level1_data['elements']不变

    # 2. 对每个分类内的元素排序
    for level1_data in level1_groups.values():
        # 排序一级分类直接包含的元素
        if level1_data['elements']:
            level1_data['elements'] = sort_elements_by_coverage_and_frequency(level1_data['elements'])

        # 排序每个二级分类的元素
        for level2_name in level1_data['level2_groups']:
            level1_data['level2_groups'][level2_name] = sort_elements_by_coverage_and_frequency(
                level1_data['level2_groups'][level2_name]
            )

    # 3. 计算每个一级分类的统计信息用于排序
    level1_scores = {}
    for level1_name, level1_data in level1_groups.items():
        # 收集该一级分类下的所有元素（包括二级分类下的）
        all_elements = level1_data['elements'][:]
        for level2_elements in level1_data['level2_groups'].values():
            all_elements.extend(level2_elements)

        if not all_elements:
            level1_scores[level1_name] = (0.0, 0, 0.0)
            continue

        # 计算统计指标
        avg_coverage = sum((e.get('共性分析') or {}).get('段落覆盖率', 0.0) for e in all_elements) / len(all_elements)
        avg_frequency = sum((e.get('共性分析') or {}).get('出现频次', 0) for e in all_elements) / len(all_elements)
        avg_intent_count = sum(calculate_intent_support_count(e) for e in all_elements) / len(all_elements)

        level1_scores[level1_name] = (avg_coverage, avg_frequency, avg_intent_count)

    # 4. 对一级分类排序
    sorted_level1 = sorted(
        level1_scores.keys(),
        key=lambda c: (-level1_scores[c][0], -level1_scores[c][1], -level1_scores[c][2])
    )

    # 5. 对每个一级分类内的二级分类排序
    for level1_name in sorted_level1:
        level1_data = level1_groups[level1_name]
        level2_groups = level1_data['level2_groups']

        if not level2_groups:
            continue

        # 计算二级分类的统计信息
        level2_scores = {}
        for level2_name, level2_elements in level2_groups.items():
            if not level2_elements:
                level2_scores[level2_name] = (0.0, 0, 0.0)
                continue

            avg_coverage = sum((e.get('共性分析') or {}).get('段落覆盖率', 0.0) for e in level2_elements) / len(level2_elements)
            avg_frequency = sum((e.get('共性分析') or {}).get('出现频次', 0) for e in level2_elements) / len(level2_elements)
            avg_intent_count = sum(calculate_intent_support_count(e) for e in level2_elements) / len(level2_elements)

            level2_scores[level2_name] = (avg_coverage, avg_frequency, avg_intent_count)

        # 排序二级分类
        sorted_level2_names = sorted(
            level2_scores.keys(),
            key=lambda c: (-level2_scores[c][0], -level2_scores[c][1], -level2_scores[c][2])
        )

        # 重新组织为有序字典
        sorted_level2_groups = {name: level2_groups[name] for name in sorted_level2_names}
        level1_data['level2_groups'] = sorted_level2_groups

    # 6. 返回排序后的结构
    return {level1_name: level1_groups[level1_name] for level1_name in sorted_level1}


def render_element_item(element: Dict[str, Any], all_elements: List[Dict[str, Any]] = None) -> str:
    """渲染单个元素项的HTML（支持详情展开，兼容新旧数据结构）

    Args:
        element: 元素数据
        all_elements: 同组所有元素（用于计算主导因素）
    """
    elem_id = element.get('id', '')
    name = element.get('名称') or ''  # 处理None的情况
    description = element.get('描述') or ''  # 处理None的情况

    # 获取类型和维度（兼容新旧结构）
    dimension = element.get('维度', {})
    if isinstance(dimension, dict):
        elem_type = dimension.get('一级', '')
        elem_type_level2 = dimension.get('二级', '')
    else:
        elem_type = element.get('类型', '')
        elem_type_level2 = ''

    # 获取分类（兼容新旧结构）
    category_data = element.get('分类', '')
    if isinstance(category_data, dict):
        category_level1 = category_data.get('一级分类', '')
        category_level2 = category_data.get('二级分类', '')
        category = get_element_category(element)
    elif isinstance(category_data, list):
        category_level1 = category_data[0] if len(category_data) > 0 else ''
        category_level2 = category_data[1] if len(category_data) > 1 else ''
        category = get_element_category(element)
    else:
        category = category_data
        category_level1 = ''
        category_level2 = ''

    category_def = element.get('分类定义', '')
    # 获取共性分析（防止为None）
    commonality = element.get('共性分析') or {}
    coverage = commonality.get('段落覆盖率', 0.0)
    frequency = commonality.get('出现频次', 0)
    paragraphs_list = commonality.get('出现段落列表', [])
    source = element.get('来源', [])
    intent_count = calculate_intent_support_count(element)
    intent_support = element.get('意图支撑', {})

    # 检查是否有详细信息
    has_details = bool(elem_type or category or category_def or paragraphs_list or source or intent_support)

    # 计算主导因素
    dominant_factor = 'coverage'  # 默认
    if all_elements:
        dominant_factor = determine_dominant_factor(element, all_elements)

    # 根据主导因素确定边框颜色
    border_color_class = f'dominant-{dominant_factor}'

    html = f'<li class="element-item {border_color_class}" data-elem-id="{elem_id}">\n'
    html += '<div class="element-header" onclick="toggleElementDetails(this)">\n'

    # 添加展开/收起图标
    if has_details:
        html += '<span class="element-toggle-icon">▶</span>\n'

    # 显示ID和名称
    if elem_id:
        html += f'<span class="element-id">#{elem_id}</span>\n'
    html += f'<span class="element-name">{html_module.escape(name)}</span>\n'

    # 显示统计指标（根据主导因素高亮）
    # 判断是否为形式元素
    is_form = isinstance(dimension, dict) and dimension.get('一级') == '形式'

    html += '<div class="element-stats">\n'

    if is_form:
        # 形式元素：显示最高相似度（基于多维度评分）
        html += f'<span class="stat-badge stat-intent">最高相似度: {intent_count:.2f}</span>\n'
    else:
        # 实质元素显示全部三个指标
        coverage_highlight = 'stat-highlight' if dominant_factor == 'coverage' else ''
        frequency_highlight = 'stat-highlight' if dominant_factor == 'frequency' else ''
        intent_highlight = 'stat-highlight' if dominant_factor == 'intent_support' else ''

        html += f'<span class="stat-badge stat-coverage {coverage_highlight}">覆盖率: {coverage:.2%}</span>\n'
        html += f'<span class="stat-badge stat-frequency {frequency_highlight}">频次: {frequency}</span>\n'
        html += f'<span class="stat-badge stat-intent {intent_highlight}">意图支撑: {intent_count}</span>\n'

    html += '</div>\n'

    html += '</div>\n'

    # 描述（始终显示）
    if description:
        html += f'<div class="element-description">{html_module.escape(description)}</div>\n'

    # 详细信息（可展开）
    if has_details:
        html += '<div class="element-details collapsed">\n'

        # 维度/类型 - 已移除，不再展示
        # if elem_type:
        #     html += '<div class="detail-section">\n'
        #     html += '<strong>维度：</strong>\n'
        #     html += '<div class="detail-content">\n'
        #     html += f'<span class="detail-tag dimension-level1">{html_module.escape(elem_type)}</span>\n'
        #     if elem_type_level2:
        #         html += f'<span class="detail-tag dimension-level2">{html_module.escape(elem_type_level2)}</span>\n'
        #     html += '</div>\n'
        #     html += '</div>\n'

        # 分类 - 已移除，不再展示
        # if category_level1 or category:
        #     html += '<div class="detail-section">\n'
        #     html += '<strong>分类：</strong>\n'
        #     html += '<div class="detail-content">\n'
        #     if category_level1:
        #         html += f'<span class="detail-tag category-level1">{html_module.escape(category_level1)}</span>\n'
        #         if category_level2:
        #             html += f'<span class="detail-tag category-level2">{html_module.escape(category_level2)}</span>\n'
        #     else:
        #         html += f'<span class="detail-tag">{html_module.escape(str(category))}</span>\n'
        #     html += '</div>\n'
        #     html += '</div>\n'

        # 分类定义
        if category_def:
            html += '<div class="detail-section">\n'
            html += '<strong>分类定义：</strong>\n'
            html += f'<div class="detail-text">{html_module.escape(category_def)}</div>\n'
            html += '</div>\n'

        # 针对"形式"维度，显示"支撑"、"推理"和"多维度评分"
        if isinstance(dimension, dict) and dimension.get('一级') == '形式':
            # 支撑
            zhicheng = element.get('支撑')
            if zhicheng:
                html += '<div class="detail-section">\n'
                html += '<strong>支撑：</strong>\n'
                html += '<div class="detail-content">\n'
                if isinstance(zhicheng, list):
                    for item in zhicheng:
                        if isinstance(item, dict):
                            item_id = item.get('id', '')
                            item_name = item.get('名称', '')
                            html += f'<span class="detail-tag">{html_module.escape(f"{item_id}: {item_name}")}</span>\n'
                        else:
                            html += f'<span class="detail-tag">{html_module.escape(str(item))}</span>\n'
                elif isinstance(zhicheng, dict):
                    # 支撑可能是对象（包含具体元素、具象概念等）
                    for key, values in zhicheng.items():
                        if isinstance(values, list):
                            html += f'<div class="detail-tag category-level1">{html_module.escape(key)}</div>\n'
                            for item in values:
                                if isinstance(item, dict):
                                    item_id = item.get('id', '')
                                    item_name = item.get('名称', '')
                                    html += f'<span class="detail-tag">{html_module.escape(f"{item_id}: {item_name}")}</span>\n'
                                else:
                                    html += f'<span class="detail-tag">{html_module.escape(str(item))}</span>\n'
                        else:
                            html += f'<span class="detail-tag">{html_module.escape(str(values))}</span>\n'
                else:
                    html += f'<span class="detail-tag">{html_module.escape(str(zhicheng))}</span>\n'
                html += '</div>\n'
                html += '</div>\n'

            # 推理
            tuili = element.get('推理')
            if tuili:
                html += '<div class="detail-section">\n'
                html += '<strong>推理：</strong>\n'
                html += f'<div class="detail-text">{html_module.escape(tuili)}</div>\n'
                html += '</div>\n'

            # 多维度评分（形式元素专用，简化为分类列表展示）
            multi_scores = element.get('多维度评分') or {}
            if multi_scores:
                html += '<div class="detail-section">\n'
                html += '<strong>多维度评分：</strong>\n'

                for score_type in ['灵感点', '目的点', '关键点']:
                    score_items = multi_scores.get(score_type) or []
                    if not score_items:
                        continue

                    html += f'<div class="score-type">{score_type}</div>\n'
                    html += '<ul class="score-list simple-score-list">\n'

                    for score_item in score_items:
                        if not isinstance(score_item, dict):
                            continue

                        # 兼容两种结构：
                        # 1）新结构：{"名称": "...", "相似度结果": [ {...}, ... ]}
                        # 2）旧结构：直接列表 [{"点","语义相似度","文本相似度",...}, ...]
                        similarity_results = score_item.get('相似度结果')

                        if similarity_results:
                            # 新结构：对每个相似度结果生成一条列表项
                            for sim in similarity_results or []:
                                if not isinstance(sim, dict):
                                    continue
                                point = sim.get('点', '')
                                semantic = sim.get('语义相似度', 0)
                                text_sim = sim.get('文本相似度', 0)

                                html += '<li class="score-item simple-score-item">\n'
                                if point:
                                    html += f'<span class="score-point-name">{html_module.escape(point)}</span>\n'
                                html += f'<span class="score-badge">语义 {semantic:.2f}</span>\n'
                                html += f'<span class="score-badge">文本 {text_sim:.2f}</span>\n'
                                html += '</li>\n'
                        else:
                            # 旧结构：当前条目本身就是一个点的评分
                            point = score_item.get('点', '')
                            semantic = score_item.get('语义相似度', 0)
                            text_sim = score_item.get('文本相似度', 0)

                            html += '<li class="score-item simple-score-item">\n'
                            if point:
                                html += f'<span class="score-point-name">{html_module.escape(point)}</span>\n'
                            html += f'<span class="score-badge">语义 {semantic:.2f}</span>\n'
                            html += f'<span class="score-badge">文本 {text_sim:.2f}</span>\n'
                            html += '</li>\n'

                    html += '</ul>\n'  # end score-list

                html += '</div>\n'  # end detail-section

        # 针对"隐含概念"，显示"来源"（声音特征、语气语调、BGM、音效等）和"时间范围"
        elem_type = element.get('类型', '')
        if elem_type == '隐含概念':
            # 来源（隐含概念的来源包含声音特征、语气语调、背景音乐、音效等）
            laiyuan = element.get('来源')
            if laiyuan and isinstance(laiyuan, dict):
                html += '<div class="detail-section">\n'
                html += '<strong>来源：</strong>\n'
                html += '<div class="detail-content">\n'
                for key, values in laiyuan.items():
                    if isinstance(values, list) and values:
                        html += f'<div class="detail-tag category-level1">{html_module.escape(key)}</div>\n'
                        for item in values:
                            html += f'<span class="detail-tag">{html_module.escape(str(item))}</span>\n'
                    elif values:
                        html += f'<span class="detail-tag">{html_module.escape(str(values))}</span>\n'
                html += '</div>\n'
                html += '</div>\n'

            # 时间范围
            time_range = element.get('时间范围')
            if time_range:
                html += '<div class="detail-section">\n'
                html += '<strong>时间范围：</strong>\n'
                html += '<div class="detail-content">\n'
                if isinstance(time_range, list):
                    for tr in time_range:
                        html += f'<span class="detail-tag source-tag">{html_module.escape(str(tr))}</span>\n'
                else:
                    html += f'<span class="detail-tag source-tag">{html_module.escape(str(time_range))}</span>\n'
                html += '</div>\n'
                html += '</div>\n'

        # 针对"抽象概念"（实质-抽象概念），显示"类型"、"来源"和"推理过程"
        # 注意：排除隐含概念（隐含概念有自己的显示逻辑）
        elif isinstance(dimension, dict) and dimension.get('二级') == '抽象概念':
            # 类型
            leixing = element.get('类型')
            if leixing:
                html += '<div class="detail-section">\n'
                html += '<strong>类型：</strong>\n'
                html += f'<div class="detail-text">{html_module.escape(str(leixing))}</div>\n'
                html += '</div>\n'

            # 来源（抽象概念的来源可能是复杂对象）
            laiyuan = element.get('来源')
            if laiyuan and isinstance(laiyuan, dict):
                html += '<div class="detail-section">\n'
                html += '<strong>来源：</strong>\n'
                html += '<div class="detail-content">\n'
                for key, values in laiyuan.items():
                    if isinstance(values, list):
                        html += f'<div class="detail-tag category-level1">{html_module.escape(key)}</div>\n'
                        for item in values:
                            if isinstance(item, dict):
                                item_id = item.get('id', '')
                                item_name = item.get('名称', '')
                                html += f'<span class="detail-tag">{html_module.escape(f"{item_id}: {item_name}")}</span>\n'
                            else:
                                html += f'<span class="detail-tag">{html_module.escape(str(item))}</span>\n'
                    else:
                        html += f'<span class="detail-tag">{html_module.escape(str(values))}</span>\n'
                html += '</div>\n'
                html += '</div>\n'

            # 推理过程
            tuili_guocheng = element.get('推理过程')
            if tuili_guocheng:
                html += '<div class="detail-section">\n'
                html += '<strong>推理过程：</strong>\n'
                html += f'<div class="detail-text">{html_module.escape(tuili_guocheng)}</div>\n'
                html += '</div>\n'

        else:
            if source:
                html += '<div class="detail-section">\n'
                html += '<strong>来源：</strong>\n'
                html += '<div class="detail-content">\n'
                for src in source:
                    html += f'<span class="detail-tag source-tag">{html_module.escape(str(src))}</span>\n'
                html += '</div>\n'
                html += '</div>\n'

        # 上下文验证（适用于具象概念）
        context_verification = element.get('上下文验证')
        if context_verification:
            html += '<div class="detail-section">\n'
            html += '<strong>上下文验证：</strong>\n'
            html += '<div class="context-verification" style="margin-top: 8px; padding: 12px; background-color: #f8f9fa; border-radius: 4px; border-left: 3px solid #6c757d;">\n'

            # 原文位置
            original_position = context_verification.get('原文位置', '')
            if original_position:
                html += '<div class="context-item" style="margin-bottom: 8px;">\n'
                html += '<strong style="color: #495057; font-size: 13px;">原文位置：</strong>\n'
                html += f'<div style="margin-top: 4px; padding: 6px 10px; background-color: #fff; border-radius: 3px; color: #212529; font-style: italic;">{html_module.escape(original_position)}</div>\n'
                html += '</div>\n'

            # 语法成分
            grammar_component = context_verification.get('语法成分', '')
            if grammar_component:
                html += '<div class="context-item" style="margin-bottom: 8px;">\n'
                html += '<strong style="color: #495057; font-size: 13px;">语法成分：</strong>\n'
                html += f'<span style="margin-left: 8px; padding: 3px 10px; background-color: #e7f3ff; color: #0066cc; border-radius: 3px; font-size: 12px;">{html_module.escape(grammar_component)}</span>\n'
                html += '</div>\n'

            # 语境判断
            context_judgment = context_verification.get('语境判断', '')
            if context_judgment:
                html += '<div class="context-item" style="margin-bottom: 0;">\n'
                html += '<strong style="color: #495057; font-size: 13px;">语境判断：</strong>\n'
                html += f'<div style="margin-top: 4px; padding: 8px 10px; background-color: #fff; border-radius: 3px; color: #495057; line-height: 1.6;">{html_module.escape(context_judgment)}</div>\n'
                html += '</div>\n'

            html += '</div>\n'
            html += '</div>\n'

        # 出现段落
        if paragraphs_list:
            html += '<div class="detail-section">\n'
            html += '<strong>出现段落：</strong>\n'
            html += '<div class="paragraphs-detail-list">\n'
            for para in paragraphs_list:
                if isinstance(para, dict):
                    # 新结构：对象包含段落ID和如何体现
                    para_id = para.get('段落ID', '')
                    how = para.get('如何体现', '')
                    html += '<div class="paragraph-detail-item">\n'
                    html += f'<span class="detail-tag para-id-tag">{html_module.escape(para_id)}</span>\n'
                    if how:
                        html += f'<div class="para-how">{html_module.escape(how)}</div>\n'
                    html += '</div>\n'
                else:
                    # 旧结构：字符串
                    html += f'<span class="detail-tag">{html_module.escape(str(para))}</span>\n'
            html += '</div>\n'
            html += '</div>\n'

        # 意图支撑
        if intent_support:
            html += '<div class="detail-section">\n'
            html += '<strong>意图支撑：</strong>\n'

            for point_type in ['灵感点', '目的点', '关键点']:
                if point_type in intent_support and intent_support[point_type]:
                    html += f'<div class="score-type">{point_type}</div>\n'
                    html += '<div class="score-list">\n'

                    for item in intent_support[point_type]:
                        point = item.get('点', '')
                        point_intention = item.get('点的意图', '')
                        support_reason = item.get('支撑理由', '')

                        html += '<div class="score-item">\n'
                        html += f'<div class="score-point">{html_module.escape(point)}</div>\n'

                        # 显示点的意图
                        if point_intention:
                            html += '<div class="point-intention">\n'
                            html += f'<strong style="color: #666;">点的意图：</strong>{html_module.escape(point_intention)}\n'
                            html += '</div>\n'

                        # 显示支撑理由
                        if support_reason:
                            html += '<div class="score-reasons">\n'
                            html += f'<strong style="color: #666;">支撑理由：</strong>\n'
                            html += f'<div class="score-reason">{html_module.escape(support_reason)}</div>\n'
                            html += '</div>\n'

                        html += '</div>\n'

                    html += '</div>\n'

            html += '</div>\n'

        html += '</div>\n'

    html += '</li>\n'
    return html


def generate_tab3_content(data: Dict[str, Any]) -> str:
    """生成Tab3内容：按层次展示（实质/形式 → 具体元素/具体概念/抽象概念 → 树形展示）"""
    html = '<div class="tab-content" id="tab3" style="display:none;">\n'

    # 添加全局控制按钮
    html += '<div class="global-controls">\n'
    html += '    <div class="color-legend">\n'
    html += '        <span class="legend-title">颜色图例（主导因素）：</span>\n'
    html += '        <span class="legend-item legend-coverage">覆盖率</span>\n'
    html += '        <span class="legend-item legend-frequency">频次</span>\n'
    html += '        <span class="legend-item legend-intent">意图支撑</span>\n'
    html += '    </div>\n'
    html += '    <button class="control-btn" onclick="toggleAllLevels(true)">全部展开</button>\n'
    html += '    <button class="control-btn" onclick="toggleAllLevels(false)">全部收起</button>\n'
    html += '</div>\n'

    if '脚本理解' in data:
        script = data['脚本理解']
        # 尝试获取元素列表，如果不存在则合并实质列表和形式列表
        elements = script.get('元素列表', [])
        if not elements:
            substance_list = script.get('实质列表', [])
            form_list = script.get('形式列表', [])
            elements = substance_list + form_list

        # 第一层：按维度.一级分组（实质 vs 形式）
        level1_groups = {}
        for elem in elements:
            dimension = elem.get('维度', {})
            if isinstance(dimension, dict):
                level1 = dimension.get('一级', '实质')
            else:
                # 兼容旧结构
                level1 = elem.get('类型', '实质')

            if level1 not in level1_groups:
                level1_groups[level1] = []
            level1_groups[level1].append(elem)

        # 按顺序渲染：实质、形式
        for level1_name in ['实质', '形式']:
            if level1_name not in level1_groups:
                continue

            level1_elements = level1_groups[level1_name]
            html += '<div class="section level1-section">\n'
            html += f'<div class="level1-header collapsible" onclick="toggleLevel1(this)">\n'
            html += '<span class="level-toggle-icon">▼</span>\n'
            html += f'<h2 class="level1-title">{level1_name} ({len(level1_elements)}个)</h2>\n'
            html += '</div>\n'
            html += '<div class="level1-content">\n'

            # 第二层：按维度.二级分组
            level2_groups = {}
            for elem in level1_elements:
                dimension = elem.get('维度', {})
                elem_type = elem.get('类型', '')
                
                # 隐含概念：优先通过类型判断（因为维度二级可能是"隐含概念"或"抽象概念"）
                if elem_type == '隐含概念':
                    level2 = '隐含概念'
                elif isinstance(dimension, dict):
                    level2 = dimension.get('二级', '具体元素')
                else:
                    # 兼容旧结构
                    elem_type_old = elem.get('类型', '实质')
                    if elem_type_old == '实质':
                        level2 = '具体元素'
                    elif elem_type_old == '具象概念':
                        level2 = '具体概念'
                    else:
                        level2 = '抽象概念'

                if level2 not in level2_groups:
                    level2_groups[level2] = []
                level2_groups[level2].append(elem)

            # 根据一级维度确定二级维度遍历顺序
            if level1_name == '实质':
                level2_order = ['具体元素', '具象概念', '隐含概念', '抽象概念']
            else:  # 形式
                level2_order = ['具体元素形式', '具象概念形式', '整体形式']

            # 按顺序渲染二级维度
            for level2_name in level2_order:
                if level2_name not in level2_groups:
                    continue

                level2_elements = level2_groups[level2_name]
                html += '<div class="level2-section">\n'
                html += f'<div class="level2-header collapsible" onclick="toggleLevel2(this)">\n'
                html += '<span class="level-toggle-icon">▼</span>\n'
                html += f'<h3 class="level2-title">{level2_name} ({len(level2_elements)}个)</h3>\n'
                html += '</div>\n'
                html += '<div class="level2-content">\n'

                # 第三层：按树形分类结构组织
                hierarchical_categories = group_elements_by_hierarchical_category(level2_elements)

                # 渲染树形分类结构
                for cat_level1_name, cat_level1_data in hierarchical_categories.items():
                    # 收集该一级分类下的所有元素
                    all_cat_elements = cat_level1_data['elements'][:]
                    for level2_elems in cat_level1_data['level2_groups'].values():
                        all_cat_elements.extend(level2_elems)

                    if not all_cat_elements:
                        continue

                    # 计算一级分类的统计信息
                    avg_coverage = sum((e.get('共性分析') or {}).get('段落覆盖率', 0.0) for e in all_cat_elements) / len(all_cat_elements)
                    avg_intent_count = sum(calculate_intent_support_count(e) for e in all_cat_elements) / len(all_cat_elements)

                    html += '<div class="category-group collapsible">\n'
                    html += '<div class="category-header" onclick="toggleCategoryGroup(this)">\n'
                    html += '<span class="category-toggle-icon">▼</span>\n'
                    html += f'<h4 class="category-title">{html_module.escape(cat_level1_name)} ({len(all_cat_elements)}个)</h4>\n'
                    html += '<div class="category-stats">\n'
                    html += f'<span class="stat-badge">平均覆盖率: {avg_coverage:.2%}</span>\n'
                    html += f'<span class="stat-badge">平均意图支撑: {avg_intent_count:.1f}</span>\n'
                    html += '</div>\n'
                    html += '</div>\n'

                    html += '<div class="category-content">\n'

                    # 渲染一级分类直接包含的元素
                    if cat_level1_data['elements']:
                        html += '<ul class="element-list">\n'
                        for elem in cat_level1_data['elements']:
                            html += render_element_item(elem, all_cat_elements)
                        html += '</ul>\n'

                    # 渲染二级分类
                    for cat_level2_name, cat_level2_elements in cat_level1_data['level2_groups'].items():
                        if not cat_level2_elements:
                            continue

                        # 计算二级分类的统计信息
                        avg_coverage_l2 = sum((e.get('共性分析') or {}).get('段落覆盖率', 0.0) for e in cat_level2_elements) / len(cat_level2_elements)
                        avg_intent_count_l2 = sum(calculate_intent_support_count(e) for e in cat_level2_elements) / len(cat_level2_elements)

                        html += '<div class="subcategory-group collapsible">\n'
                        html += '<div class="subcategory-header" onclick="toggleSubcategoryGroup(this)">\n'
                        html += '<span class="subcategory-toggle-icon">▼</span>\n'
                        html += f'<h5 class="subcategory-title">{html_module.escape(cat_level2_name)} ({len(cat_level2_elements)}个)</h5>\n'
                        html += '<div class="subcategory-stats">\n'
                        html += f'<span class="stat-badge-small">覆盖率: {avg_coverage_l2:.2%}</span>\n'
                        html += '</div>\n'
                        html += '</div>\n'

                        html += '<div class="subcategory-content">\n'
                        html += '<ul class="element-list">\n'
                        for elem in cat_level2_elements:
                            html += render_element_item(elem, cat_level2_elements)
                        html += '</ul>\n'
                        html += '</div>\n'
                        html += '</div>\n'

                    html += '</div>\n'
                    html += '</div>\n'

                html += '</div>\n'  # level2-content
                html += '</div>\n'  # level2-section

            html += '</div>\n'  # level1-content
            html += '</div>\n'  # level1-section

    html += '</div>\n'
    return html