基于笔者之前写的博客基础上：https://blog.csdn.net/zhanghan11366/article/details/142139488【基于开源WQ装备知识图谱的智能问答全流程构建】进行优化。新增处理基于特定格式下的WQ文档，抽取文档的WQ属性和关系，并抽取对应WQt图片存储至minio中。
1 文档格式如下：
2 提取文档中的WQ信息
- 配置如下：
import argparseclass Args:@staticmethoddef parse():parser = argparse.ArgumentParser()return parser@staticmethoddef initialize(parser):parser.add_argument('--weapon_realtion_api', default='http://0.0.0.0:6410/weapon',help='Weapons and equipment relationship extraction API interface')parser.add_argument('--word_extraction_api', default='http://1.95.39.242:2011/attribute_all',help='Parse word documents and extract weapons and equipment attributes api interface')parser.add_argument('--neo4j_url', default='bolt://localhost:7687',help='neo4j login website')parser.add_argument('--neo4j_usename', default='neo4j',help='neo4j login username')parser.add_argument('--neo4j_password', default='neo4jZH',help='neo4j login password')parser.add_argument('--unstr_file_path', default='./data/unstr/武器装备-test.docx',help='Unstructured document parsing path')parser.add_argument('--unstr_save_file_path', default='./data/unstr/word_weapon_basic_info.txt',help='The path to save the unstructured document after parsing')parser.add_argument('--weapon_input_file', default='./data/weapon/weapon_data.txt',help='Weapons and equipment relationship extraction input address')return parserdef get_parser(self):parser = self.parse()parser = self.initialize(parser)return parser.parse_args()
- 提取代码
import requests, config# 调用
args = config.Args().get_parser()def get_word_weapon(file_path):# 发送 POST 请求并上传文件with open(file_path, 'rb') as file:files = {'file': file}response = requests.post(args.word_extraction_api, files=files)# 检查响应状态码if response.status_code == 200:try:# 尝试以 JSON 格式解析响应response_json = response.json()return response_json.get('data')except ValueError:print("响应不是 JSON 格式:")print(response.text)else:print(f"请求失败，状态码: {response.status_code}")print(f"响应内容: {response.text}")def save_word_weapon_basic_info(weapon_data, file_path):# 循环每个武器数据，将其基本情况写入TXT文件with open(file_path, 'w', encoding='utf-8') as file:for weapon in weapon_data:basic_info = weapon['基本情况'].replace('\n', ' ')file.write(basic_info + '\n') # 换行区分不同武器的信息print(f"武器的基本情况已保存至 {file_path}")if __name__ == "__main__":# 文件路径weapon_data = get_word_weapon(args.unstr_file_path)save_word_weapon_basic_info(weapon_data, args.unstr_save_file_path)
- 其中接口信息如下：
3 图文匹配
抽取对应WQt图片存储至minio中，结果如下。
后续流程与https://blog.csdn.net/zhanghan11366/article/details/142139488【基于开源WQ装备知识图谱的智能问答全流程构建】一致。
