Рекомендуемый стек
Что вы создаёте? Semantic search
Одна технология на позицию, с обоснованием и тем, чем её можно заменить.
Embedding
conceptDense numeric vector representing the meaning of a piece of text.
Why editorial
Everything else in semantic search depends on the quality of the embedding: a weak one makes good ranking impossible.
Vector database
databaseOpen-source vector database with hybrid keyword and vector search built in.
Альтернативы
Base model
modelEncoder-only Transformer pre-trained with masked language modelling.
Альтернативы
Reranking
techniqueSecond-stage ordering of retrieved candidates by a more accurate scoring model.
Why выведено из графа
Retrieve-then-rerank is the standard shape of a semantic search pipeline.