📄 中文摘要
在过去一个季度，我们团队面临着会议记录的困扰，每周有超过40场会议，涉及12人，行动项分散在Slack、电子邮件和Google文档中。由于任务常常被埋没在2000字的转录文本中，难免有人错过截止日期。为了解决这个问题，我构建了一个系统，可以实时监听会议，提取结构化任务，并将其分发给相关人员。该系统已在生产环境中运行6个月，每月处理约200场会议。系统的核心在于如何有效地进行转录和任务提取，而不仅仅依赖于简单的转录工具和LLM。
Whisper + LLM 任务提取：我的会议智能架构
出处: Whisper + LLM Task Extraction: My Meeting Intelligence Architecture
发布: 2026年3月24日
📄 English Summary
Whisper + LLM Task Extraction: My Meeting Intelligence Architecture
Last quarter, the team faced challenges with meeting notes, holding over 40 meetings weekly among 12 members, with action items scattered across Slack, email, and Google Docs. Tasks often got buried in lengthy transcripts, leading to missed deadlines. To address this, a system was built that listens to meetings, extracts structured tasks, and routes them to the appropriate individuals. This system has been in production for six months, processing around 200 meetings monthly. The key lies in effectively transcribing and extracting tasks, rather than relying solely on naive transcription tools and LLMs.