方法论
Methodology
我们如何筛选、标注和呈现每日 FemTech 资讯
How we curate, tag, and present daily FemTech intelligence
01 数据源
01 Data Sources
我们的数据管道每日从22 个数据源抓取最新内容，覆盖三类信息通道：
- FemTech 专用信源：Femtech Insider、FemTech World 等行业媒体，全量收录。
- 通用科技/医疗信源：MedTech Dive、Medical Xpress 等泛医疗科技媒体，通过 63 个中英文关键词（涵盖女性生理、妇产科、激素、FemTech 产品等）过滤相关文章。
- 学术论文/监管：openFDA（510(k) 器械审批）、PubMed（生物医学论文）、ClinicalTrials.gov（临床试验），每日各取最近 5 条 FemTech 相关记录。
Our data pipeline scans 22 sources daily across three information channels:
- Dedicated FemTech feeds: Industry outlets like Femtech Insider and FemTech World—all articles accepted without keyword filtering.
- General med-tech feeds: Broader outlets like MedTech Dive and Medical Xpress, filtered by 63 bilingual keywords covering women's physiology, OB/GYN, hormones, and FemTech products.
- Academic & regulatory: openFDA (510(k) device clearances), PubMed (biomedical literature), ClinicalTrials.gov (clinical studies)—5 most recent FemTech-relevant records each, daily.
02 筛选标准
02 Selection Criteria
我们对每篇文章执行两层筛选：
- 关键词初筛：标题或摘要命中关键词列表中至少一个词条。关键词覆盖女性生理（月经、更年期、盆底肌）、妇产科（产科、子宫、卵巢）、FemTech 产品（吸奶器、可穿戴、月经杯）等六大类别。
- 标题去重：同一标题仅保留一篇，避免多源重复。
Every article passes through two filters:
- Keyword screening: Title or summary must match at least one entry from our keyword list—spanning women's physiology (menstruation, menopause, pelvic floor), OB/GYN (obstetrics, uterus, ovaries), and FemTech products (breast pumps, wearables, menstrual cups).
- Title deduplication: Identical titles are merged to avoid duplicates across sources.
03 标签体系
03 Tagging System
每篇文章通过 AI 自动匹配四维标签：
Each article is auto-tagged across four dimensions by AI:
contentTypetechbodylifeStagemethod标签自动匹配基于结构化关键词词库（KEYWORD_REGISTRY），按标题与摘要中的命中情况打分分配，覆盖技术、健康领域、生命周期与方法场景四个维度，规则透明、可审计。
Tags are auto-assigned via a structured keyword registry (KEYWORD_REGISTRY), scored by title and summary matches across four dimensions: technology, health domain, life stage, and methodology. The rules are transparent and auditable.
04 AI 辅助编辑流程
04 AI-Assisted Editorial Pipeline
我们搭建了一套由 Edit AI、Writer AI 与 Review AI 构成的分级多智能体编辑管线，以分步推理与交叉验证机制处理原始素材——从结构化提取、语义增强到事实核查与双语审校，每一环节均引入领域上下文约束与质量阈值。
原始数据清洗与结构化，提取标题、摘要、信源等元信息
生成中英双语标题、摘要，自动匹配标签与分类
验证事实准确性、标签一致性、中英翻译质量
We operate a tiered multi-agent editorial pipeline composed of Edit AI, Writer AI, and Review AI, processing raw material through staged reasoning and cross-validation—from structured extraction and semantic enrichment to fact-checking and bilingual review—each layer constrained by domain context and quality thresholds, far beyond what a single prompt could deliver.
Raw data cleaning and structuring—extracts metadata: titles, summaries, sources
Generates bilingual titles and summaries, auto-assigns tags and categories
Verifies factual accuracy, tag consistency, and translation quality
05 透明度承诺
05 Transparency
- 信源可查：每篇文章均标注原始来源链接，读者可自行验证。
- 持续迭代：方法论将随管线升级而更新。
- Verifiable sources: Every article links to its original source. Readers can verify independently.
- Living document: This methodology page is updated as the pipeline evolves.