Skip to content

Table Item Embedding Refactor

The ml_feat__item_embedding_refactor table contains item embeddings produced by embedding-gemma-300m. Each item has:

  • a content hash of the item's metadata.
  • multiple embedding columns that are defined in the item_embedding microservice config.

Table description

name data_type description
item_id STRING Identifier for the item associated with the offer used internally by the data science team.
content_hash STRING Hash of item's metadata.
semantic_content_sts ARRAY Embedding of item's metadata with "STS" prompt name. Encoded with embedding-gemma-300m model.
semantic_content_clustering ARRAY Embedding of item's metadata with "Clustering" prompt name. Encoded with embedding-gemma-300m model.