🌐 金融行业对话:MongoDB Atlas矢量搜索实际用户案例(搜索内部PDF文档)💰

Danny Chan - Aug 7 - - Dev Community

用例1:
📁 使用矢量搜索和LLM的PDF搜索应用程序
💼 检索增强型生成(RAG)


问题: 🔍

  • PDF是非结构化数据
  • 难以搜索
  • 领先的理赔调查员审查合同和理赔
  • 需要高效和准确性来解决这一历史性的繁琐任务
  • 总结文件内容
  • 指示信息来源


挑战: 🤯

  • 承保人和理赔调查员要浏览大量的指南、合同和报告页面
  • 耗时
  • 导致昂贵的错误(风险估算不准确)


详情: 🗂️

  • 根据用户所在国家提供PDF
  • 矢量搜索查询特定文档
  • 过滤掉无关的PDF


索引预筛选: 🔍

  • 必须索引具有数据类型的字段
  • 只包括具有正确数据类型的记录


专用搜索节点架构: 🛠️

  • 同时支持Atlas搜索和矢量搜索工作负载



用例2:
📊 使用LLM和矢量搜索的理赔管理,用于RAG


挑战: 🙆‍♀️

  • 理赔调查员从不同格式的系统中提取和汇总信息
  • 内容共享平台
  • 客户信息被锁定在遗留的客户关系管理系统中
  • 与理赔有关的图片和语音报告
  • 非结构化数据
  • 简化操作


解决方案: 🔍🤖

  • Atlas矢量搜索
  • 大语言模型(LLM)
  • 检索增强型生成(RAG)


矢量搜索查询: 🔍

  • 获取矢量嵌入和纯文本元数据
  • 不需要检索数据


聚合管道: 🧠

  • 处理多个文档并返回计算结果


Reference:
Claim management using LLMs and vector search for RAG
https://www.mongodb.com/solutions/solutions-library/claim-management-llms-vector-search

Harnessing the Power of Atlas Vector Search and Open Source Models
https://www.mongodb.com/developer/products/atlas/building-generative-ai-applications-vector-search-open-source-models/?hideMenu=1

Build a PDF search application with vector search and LLMs
https://www.mongodb.com/solutions/solutions-library/pdf-search-with-vector-search-and-llm


Editor

Image description

Danny Chan, specialty of FSI and Serverless

Image description

Kenny Chan, specialty of FSI and Machine Learning

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Terabox Video Player