Unverified50% confidenceSolutionExact time
Entity-driven retrieval refines the matching granularity from 'entire sentence to entire name' to 'term to entity': even if a table name is an internal abbreviation, as long as that abbreviation has co-occurred with a certain business term in the query logs, a mapping can be established.
1
Sources
50%
Confidence
Long-term
Relevance
10/27/2026
First Seen
Sources
Related Entities
Related Claims
Unverified知识系统去重时应先识别不同名称是否指向同一对象，若页面已存在则整合进原页面并保留各自来源65% similarUnverifiedA hybrid strategy is recommended for product naming: AI generates descriptive semantic labels, and a numeric suffix is then automatically appended to ensure uniqueness.65% similarUnverified去重应根据标题、别名、对象类型和上下文共同判断是否重复，而非关键词匹配64% similarVerified关键词搜索基于精确字符匹配无法理解同义词，语义搜索依赖向量嵌入可捕捉语义关联，两者结合的混合检索效果优于单一方法64% similarUnverifiedCognee通过LLM的命名实体识别和关系抽取能力，自动从非结构化文本中识别实体与关系并写入图数据库，查询时利用图遍历算法进行多跳推理64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1102560API
curl https://kongchang.com/api/v1/knowledge/claims/1102560MCP
get_claim(id=1102560)