Query rewrite for low performing queries based on customer behavior

A computer-implemented method, comprising:

• receiving, by one or more processors of one or more computing devices, a product query from a user electronic device;
• determining, by the one or more processors, a query rewrite for the product query based on a query rewrite model, wherein the query rewrite model is configured to determine the query rewrite by:
• comparing the product query to vectors representing a plurality of product query arrays each comprising a plurality of individual product queries received during a single user search session, wherein text of each of the plurality of individual product queries in each product query array is treated as a whole token;
• determining one of the vectors that is closest to a product query vector;
• and
• determining a query rewrite based on the determined one of the vectors;
• determining, by the one or more processors, search results for the product query using the query rewrite;
• and
• sending, by the one or more processors to the user electronic device, information for presenting the search results on a display of the user electronic device responsive to the product query.

A computer-implemented method, comprising:

• receiving, by one or more processors of the one or more computing devices, a product query from a user electronic device;
• determining, by the one or more processors, a query rewrite for the product query by retrieving, from a lookup table, a query rewrite candidate related to the product query, wherein the lookup table is generated using a query rewrite model by:
• inputting a plurality of product query arrays each comprising a plurality of individual product queries received during a single user search session into the query rewrite model, wherein text of each of the plurality of individual product queries in each product query arrays is treated as a whole token;
• generating vectors representing the plurality of product query arrays;
• determining a query rewrite candidate for each of the vectors;
• and
• storing the vectors and each respective query rewrite candidate in the lookup table;
• determining, by the one or more processors, search results for the product query using the query rewrite;
• and
• sending, by the one or more processors to the user electronic device, information for presenting the search results on a display of the user electronic device responsive to the product query.

A non-transitory computer-readable medium having stored thereon instructions that are executable by a processor to cause a system to perform operations comprising:

• receive a product query from a user electronic device;
• determine a query rewrite for the product query based on a query rewrite model, wherein the query rewrite model is configured to determine the query rewrite by:
• compare the product query to vectors representing a plurality of product query arrays each comprising a plurality of individual product queries received during a single user search session, wherein text of each of the plurality of individual product queries in each product query array is treated as a whole token;
• determine one of the vectors that is closest to a product query vector;
• and
• determine a query rewrite based on the determined one of the vectors;
• determine search results for the product query using the query rewrite;
• and
• send, to the user electronic device, information for presenting the search results on a display of the user electronic device responsive to the product query.

A non-transitory computer-readable medium having stored thereon instructions that are executable by a processor to cause a system to perform operations comprising:

• receive a product query from a user electronic device;
• determine a query rewrite for the product query by retrieving, from a lookup table, a query rewrite candidate related to the product query, wherein the lookup table is generated using a query rewrite model by:
• input a plurality of product query arrays each comprising a plurality of individual product queries received during a single user search session into the query rewrite model, wherein text of each of the plurality of individual product queries in each product query arrays is treated as a whole token;
• generate vectors representing the plurality of product query arrays;
• determine a query rewrite candidate for each of the vectors;
• and
• store the vectors and each respective query rewrite candidate in the lookup table;
• determine, by the one or more processors, search results for the product query using the query rewrite;
• and
• sending, by the one or more processors to the user electronic device, information for presenting the search results on a display of the user electronic device responsive to the product query.