Semantic search doesn't require vector embeddings. Hierarchical taxonomies offer an alternative approach that provides better matching criteria and domain-specific categorization. Using managed vocabularies with hierarchical tokenization in BM25 indexes can achieve semantic similarity while maintaining precise

16m read time From softwaredoug.com
Post cover image
Table of contents
Semantic search - what exactly is it?Embeddings for semantic searchThe missing semantic search piece of vector retrievalThe other semantic search approach (hierarchical managed taxonomies)Semantic similarity doesn’t need a vector indexBuilding a taxonomy: start simple and evolveClassifying into taxonomies has never been easierFinal thoughts

Sort: