Solving sparse data problems in a recommendation system with cold start
AssigneeSap SE
InventorsNick Pendar, Arjun Rao
Life of this patent
Estimating the term…
Abstract
Recommendation services typically struggle with sparse data scenarios. A cold generated item start technique can find a warm generated item as a suitable proxy in a matrix representing co-occurrences of generated items. Embeddings can be used to determine distance between items. The technologies are useful for providing recommendations even in scenarios involving little or no transaction data.
Claims20
A computer-implemented method comprising:
- building a generated-item-by-generated-item matrix representing correlations between a plurality of generated items based on observed transactions for the generated items, wherein the generated-item-by-generated-item matrix comprises at least one sparse data scenario for a generated item represented in the matrix;
- finding a cold generated item in the generated-item-by-generated-item matrix;
- embedding representations of at least a subset of the plurality of generated items into an embedded vector space;
- determining a location of the cold generated item in the embedded vector space via one or more attributes of the cold generated item;
- finding, in the embedded vector space, a nearest neighbor warm generated item of the cold generated item with a nearest neighbor search technique;
- identifying the nearest neighbor warm generated item as a suitable proxy for the cold generated item during recommendation processing;
- and
- outputting a recommendation based on the generated-item-by-generated-item matrix.
A computing system comprising:
- a generated-item-by-generated-item matrix stored in one or more non-transitory computer-readable media and representing observed co-occurrences within transactions of generated items represented in the matrix;
- an internal representation of a high-dimensional embedded vector space in which a plurality of warm generated items is represented, and a cold generated item has a location;
- and
- a recommendation extension service configured to identify a warm approximate nearest neighbor of the cold generated item in the high-dimensional embedded vector space and copy entries for the warm approximate nearest neighbor in the matrix to entries for the cold generated item in the matrix.
One or more non-transitory computer-readable media having stored therein computer-executable instructions that when executed by a computing system, cause the computing system to perform:
- representing a plurality of generated items in a generated item co-occurrence matrix;
- finding, in an embedded vector space representing the generated items based on respective textual descriptions of the generated items, a close warm generated item to a cold generated item with an approximate nearest neighbor search technique;
- copying values for the close warm generated item to the cold generated item into the matrix for the cold generated item;
- receiving a request to recommend an incoming generated item;
- in the generated item co-occurrence matrix, reading a copied entry for the incoming generated item, wherein the copied entry indicates a recommended generated item;
- and
- responsive to the request, responding with the recommended generated item.
Dependencies are inferred from the claim text.