Method, computer program product, and system for selecting an item for inclusion in an order from a user of an online concierge

A computer program product comprising a non-transitory computer readable storage medium having instructions encoded thereon that, when executed by a processor of an online system, cause the processor to:

• receive a generic item description from a user, the generic item description being a search query inputted into a search field of an interface;
• retrieve a taxonomy stored by the online system, the taxonomy maintaining associations between generic item descriptions and attributes of specific items that are associated with serial numbers, the attributes comprising information of stores that offer one or more specific items;
• identify, based on the taxonomy, a set of candidate items that match the generic item description specified in the search field;
• traverse one or more paths of the taxonomy from the set of candidate items to identify one or more specific items that are associated with corresponding serial numbers;
• select the one or more specific items from the set of candidate items, wherein select the specific item comprises applying a machine learning model to rank the set of candidate items by availabilities, wherein the machine learning model is trained by:
• initiating a set of weights of the machine learning model,
• receiving a training set that comprises inventory information and past outcomes of delivery orders,
• comparing an item's availability across multiple warehouses to determine whether the item is chronically unavailable,
• updating the training set based on the item's availability across multiple warehouses, and
• adjusting the weights of the machine learning model based on the training set;
• and
• cause, responsive to the generic item description being inputted into the search field of the interface, the interface to automatically provide the one or more specific items for display without the user providing additional inputs to identify the specific item, thereby reducing a number of steps the user needs to navigate in the interface to include a specific item.

An online system comprising:

• one or more processors;
• and
• memory configured to store code comprising instructions, where the instructions, when executed by the one or more processors, cause the one or more processors to:
• receive a generic item description from a user, the generic item description being a search query inputted into a search field of an interface;
• retrieve a taxonomy stored by the online system, the taxonomy maintaining associations between generic item descriptions and attributes of specific items that are associated with serial numbers, the attributes comprising information of stores that offer one or more specific items;
• identify, based on the taxonomy, a set of candidate items that match the generic item description specified in the search field;
• traverse one or more paths of the taxonomy from the set of candidate items to identify one or more specific items that are associated with corresponding serial numbers;
• select the one or more specific items from the set of candidate items, wherein select the specific item comprises applying a machine learning model to rank the set of candidate items by availabilities, wherein the machine learning model is trained by:
• initiating a set of weights of the machine learning model,
• receiving a training set that comprises inventory information and past outcomes of delivery orders,
• comparing an item's availability across multiple warehouses to determine whether the item is chronically unavailable,
• updating the training set based on the item's availability across multiple warehouses, and
• adjusting the weights of the machine learning model based on the training set;
• and
• cause, responsive to the generic item description being inputted into the search field of the interface, the interface to automatically provide the one or more specific items for display without the user providing additional inputs to identify the specific item, thereby reducing a number of steps the user needs to navigate in the interface to include a specific item.