revision/translate_specific_gexf.py

import xml.etree.ElementTree as ET
from xml.dom import minidom
import re

def create_specific_translation_dict():
    """
    基于实际GEXF文件内容创建专门的翻译词典
    """
    translation_dict = {
        # 从GEXF文件中提取的实际术语翻译
        
        # 岩石类型 - Rock Types
        "粉砂岩": "Siltstone",
        "灰岩": "Limestone", 
        "板岩": "Slate",
        "砂岩": "Sandstone",
        "火山岩": "Volcanic rock",
        "二云石英片岩": "Two-mica quartz schist",
        "岩石": "Rock",
        "硅质岩": "Chert",
        "黑云斜长片麻岩": "Biotite plagioclase gneiss",
        "长石石英砂岩": "Feldspathic quartz sandstone",
        "长石砂岩": "Feldspathic sandstone", 
        "岩屑砂岩": "Lithic sandstone",
        "粉砂质板岩": "Silty slate",
        "泥晶灰岩": "Micritic limestone",
        "碎屑岩": "Clastic rock",
        "碳酸盐岩": "Carbonate rock",
        "俄巴达动灰岩": "Obadong Limestone",
        "安山岩": "Andesite",
        "安山玄武岩": "Andesitic basalt",
        "达龙砂岩": "Dalong Sandstone",
        "英安岩": "Dacite",
        
        # 地层单位和复合岩石名称
        "九十道班组": "Jiushidaoban Formation",
        "开心岭群": "Kaixinling Group",
        "结隆组": "Jielong Formation",
        "多彩蛇绿混杂岩": "Duocai Ophiolite Complex",
        "通天河蛇绿混杂岩": "Tongtianhe Ophiolite Complex",
        "波里拉组": "Bolila Formation",
        "诺日巴尕日保组": "Nuoribagaribao Formation",
        "含煤碎屑岩组": "Coal-bearing clastic formation",
        "结扎群": "Jiezha Group",
        "尕笛考组": "Gadikao Formation",
        "甲丕拉组": "Jiapila Formation",
        "巴颜喀拉山群": "Bayankala Group",
        "灰色灰绿色厚层中－细粒岩屑长石砂岩": "Gray to grayish-green thick-bedded medium- to fine-grained lithic arkose",
        "碎屑岩组": "Clastic formation",
        "查涌蛇绿混杂岩": "Chayong Ophiolite Complex",
        "杂多群": "Zaduo Group",
        "测区南部岩石": "Southern survey area rocks",
        "岩石组合": "Rock association",
        "杂多群碎屑岩组": "Zaduo Group clastic formation",
        "巴塘群": "Batang Group",
        "古中元古代宁多岩群": "Paleo-Mesoproterozoic Ningduo Group",
        "格仁火山岩": "Geren Volcanics",
        "晚三叠世巴塘群及结扎群": "Late Triassic Batang Group and Jiezha Group",
        "上段": "Upper member",
        "火山岩组": "Volcanic formation",
        "晚三叠世巴塘群": "Late Triassic Batang Group",
        "西金乌兰－金沙江地层分区": "Xijinwulan-Jinsha River stratigraphic division",
        "该套地层": "This stratigraphic sequence",
        "新近纪曲果组": "Neogene Quguo Formation",
        "相似地层": "Correlative strata",
        "含矽线红柱石黑云母斜长片麻岩": "Sillimanite-andalusite-bearing biotite-plagioclase gneiss",
        "沱沱河组": "Tuotuohe Formation",
        "砂岩组（2-3B）": "Sandstone formation (2-3B)",
        "晚三叠世结扎群": "Late Triassic Jiezha Group",
        "巴贡组": "Bagong Formation",
        "测区结扎群": "Survey area Jiezha Group",
        "宁多群": "Ningduo Group",
        "青海省第二区调队": "Qinghai Province Second Geological Survey Team",
        "三叠纪查涌蛇绿混杂岩": "Triassic Chayong Ophiolite Complex",
        "早中二叠世尕笛考组": "Early-Middle Permian Gadikao Formation",
        "晚二叠世－早三叠世火山岩组": "Late Permian-Early Triassic volcanic formation",
        "当江荣火山岩": "Dangjianglong Volcanics",
        
        # 矿物成分
        "斑晶": "Phenocryst",
        "斜长石": "Plagioclase",
        "石英": "Quartz",
        "黑云母": "Biotite",
        "黄铁矿晶体": "Pyrite crystals",
        "砾石": "Gravel",
        
        # 古生物
        "双壳类": "Bivalvia",
        "腕足类": "Brachiopoda", 
        "头足类": "Cephalopoda",
        
        # 地理位置
        "测区": "Study area",
        "青海省": "Qinghai Province",
        
        # 地质年代
        "晚三叠世": "Late Triassic",
        "早中二叠世": "Early-Middle Permian",
        "三叠纪": "Triassic",
        "早石炭世": "Early Carboniferous",
        "二叠纪": "Permian",
        
        # 颜色和描述
        "灰绿色": "Grayish green",
        
        # 构造特征
        "片状构造": "Foliated structure",
        "斑状结构": "Porphyritic texture",
        "水平层理": "Horizontal bedding",
        
        # 地质关系
        "岩性": "Lithology",
        "出露于": "Exposed", 
        "出露地层": "Outcrop",
        "含有": "Contained",
        "古生物": "Paleontological",
        "属于": "Belongs to",

        "断层接触": "Fault contact",
        "包含": "Includes",
        "所属年代": "Age",
        "发育": "Develop",
        "角度不整合": "Angular unconformity",
        "命名": "Named by",
        "创名": "Established by",
        "结构": "Structure",
        "厚度": "Thickness",

        "分布形态": "Distribution pattern",
        "坐标": "Coordinates",
        "整合接触": "Integration of contacts",
        "不整合接触": "Unconformity contact",
        "海拔": "Elevation",
        "位于": "Located at",
        "面积": "Exposed area",
        "大地构造位置": "Geotectonic position",
        "地层区划": "Stratigraphical zoning",
        "假整合接触": "Consolidated contact",
        "行政区划": "Administrative area",
        "吞噬": "Engulfed",
        "长度": "Length",
        "侵入": "Invade",




        # 重要性等级
        "高": "High",
        "中": "Medium",
        "低": "Low"
    }
    
    return translation_dict

def translate_geological_term_specific(chinese_term, translation_dict):
    """
    翻译特定的地质术语
    """
    # 直接匹配
    if chinese_term in translation_dict:
        return translation_dict[chinese_term]
    
    # 如果找不到精确匹配，返回原文
    return chinese_term

def translate_specific_gexf_file(input_file, output_file=None):
    """
    翻译特定GEXF文件中的中文地质术语为英文
    
    参数:
    - input_file: 输入的GEXF文件路径  
    - output_file: 输出的GEXF文件路径，如果为None则在原文件名后加_english
    """
    if output_file is None:
        output_file = input_file.replace('.gexf', '_english.gexf')
    
    # 获取翻译词典
    translation_dict = create_specific_translation_dict()
    
    try:
        # 读取并解析GEXF文件
        tree = ET.parse(input_file)
        root = tree.getroot()
        
        # 统计翻译数量
        translated_nodes = 0
        translated_edges = 0
        untranslated_terms = set()
        
        print(f"开始处理GEXF文件: {input_file}")
        
        # 翻译节点标签
        nodes = root.findall('.//{http://www.gexf.net/1.2draft}node')
        print(f"找到 {len(nodes)} 个节点")
        
        for node in nodes:
            original_label = node.get('label', '')
            if original_label:
                translated_label = translate_geological_term_specific(original_label, translation_dict)
                if translated_label != original_label:
                    node.set('label', translated_label)
                    translated_nodes += 1
                    print(f"节点翻译: {original_label} -> {translated_label}")
                else:
                    untranslated_terms.add(original_label)
        
        # 翻译边标签和属性值
        edges = root.findall('.//{http://www.gexf.net/1.2draft}edge')
        print(f"找到 {len(edges)} 条边")
        
        for edge in edges:
            # 翻译边的label属性
            original_edge_label = edge.get('label', '')
            if original_edge_label:
                translated_edge_label = translate_geological_term_specific(original_edge_label, translation_dict)
                if translated_edge_label != original_edge_label:
                    edge.set('label', translated_edge_label)
                    print(f"边标签翻译: {original_edge_label} -> {translated_edge_label}")
                else:
                    untranslated_terms.add(original_edge_label)
            
            # 翻译边的属性值
            for attvalue in edge.findall('.//{http://www.gexf.net/1.2draft}attvalue'):
                attr_for = attvalue.get('for', '')
                original_value = attvalue.get('value', '')
                
                if attr_for in ['label', 'relationship'] and original_value:
                    translated_value = translate_geological_term_specific(original_value, translation_dict)
                    if translated_value != original_value:
                        attvalue.set('value', translated_value)
                        translated_edges += 1
                    else:
                        untranslated_terms.add(original_value)
        
        # 翻译节点的重要性属性值
        importance_attrs = root.findall('.//{http://www.gexf.net/1.2draft}attvalue[@for="importance"]')
        print(f"找到 {len(importance_attrs)} 个重要性属性")
        
        for attvalue in importance_attrs:
            original_importance = attvalue.get('value', '')
            if original_importance:
                translated_importance = translate_geological_term_specific(original_importance, translation_dict)
                if translated_importance != original_importance:
                    attvalue.set('value', translated_importance)
                    print(f"重要性翻译: {original_importance} -> {translated_importance}")
        
        # 更新meta信息
        meta = root.find('.//{http://www.gexf.net/1.2draft}meta')
        if meta is not None:
            # 更新描述
            description = meta.find('.//{http://www.gexf.net/1.2draft}description')
            if description is not None:
                description.text = "Geological Knowledge Graph - English Translation"
            
            # 更新创建者
            creator = meta.find('.//{http://www.gexf.net/1.2draft}creator')
            if creator is not None:
                creator.text = "Geological Knowledge Graph Translator"
        
        # 保存翻译后的文件
        rough_string = ET.tostring(root, encoding='unicode')
        reparsed = minidom.parseString(rough_string)
        pretty_xml = reparsed.toprettyxml(indent="  ", encoding=None)
        
        # 移除空行
        pretty_lines = [line for line in pretty_xml.split('\n') if line.strip()]
        pretty_xml = '\n'.join(pretty_lines)
        
        with open(output_file, 'w', encoding='utf-8') as f:
            f.write(pretty_xml)
        
        # 输出翻译统计
        print(f"\nGEXF英文翻译完成: {output_file}")
        print(f"翻译的节点数: {translated_nodes}")
        print(f"翻译的边关系数: {translated_edges}")
        print(f"翻译词典包含术语: {len(translation_dict)} 个")
        
        if untranslated_terms:
            print(f"\n保持原文的术语 ({len(untranslated_terms)} 个):")
            for term in sorted(untranslated_terms):
                print(f"  - {term}")
                
            # 保存未翻译术语到文件
            untranslated_file = output_file.replace('.gexf', '_untranslated_terms.txt')
            with open(untranslated_file, 'w', encoding='utf-8') as f:
                f.write("未翻译的地质术语列表:\n")
                f.write("=" * 50 + "\n")
                for term in sorted(untranslated_terms):
                    f.write(f"{term}\n")
            print(f"未翻译术语已保存到: {untranslated_file}")
        else:
            print("所有术语均已成功翻译!")
        
        print("\n翻译特点:")
        print("- 专门针对此GEXF文件中的术语进行翻译")
        print("- 保持了地质术语的专业性和准确性")
        print("- 纯英文输出，适合国际学术交流")
        print("- 保留了原有的网络结构和可视化属性")
        
        return output_file
        
    except Exception as e:
        print(f"翻译GEXF文件时出错: {e}")
        import traceback
        traceback.print_exc()
        return None

# 主函数
def main():
    """主函数，执行翻译任务"""
    
    input_file = 'knowledge_graph.gexf'
    
    print("开始翻译地质知识图谱GEXF文件...")
    print("=" * 60)
    
    # 检查文件是否存在
    import os
    if not os.path.exists(input_file):
        print(f"错误：找不到输入文件 {input_file}")
        print("请确保 knowledge_graph.gexf 文件在当前目录中")
        return
    
    # 创建纯英文版本
    english_file = translate_specific_gexf_file(input_file, 'knowledge_graph_english.gexf')
    
    print("\n" + "=" * 60)
    print("翻译任务完成!")
    if english_file:
        print(f"输出文件: {english_file}")

if __name__ == "__main__":
    main()