Unverified50% confidenceSolutionExact time
Athena segments conversations into multiple "chains" by topic by computing the cosine similarity between adjacent conversation turns, closing a chain when the similarity drops.
1
Sources
50%
Confidence
Long-term
Relevance
9/24/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSentence-BERT通过孪生网络结构将句子映射为固定维度的稠密向量，专为句子级语义相似度优化60% similarUnverifiedSynonym normalization can be achieved by using sentence embedding models such as Sentence-BERT to map words into dense vectors, and then computing semantic proximity via cosine similarity.59% similarUnverified语义分块基于句子嵌入的相似度变化检测段落边界，可避免固定长度分块导致的语义断裂59% similarUnverifiedAI Search采用递归式切分方式进行文本切分，优先在段落或句子的自然边界进行拆分，并通过Overlap重叠机制让相邻切片保留重叠文本58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/947109API
curl https://kongchang.com/api/v1/knowledge/claims/947109MCP
get_claim(id=947109)