CPC G06F 16/957 (2019.01) [G06F 16/9035 (2019.01); G06F 16/904 (2019.01); G06F 16/9535 (2019.01); G06F 16/9538 (2019.01); G06F 16/954 (2019.01)] | 18 Claims |
1. A method for generating an entity list, the method comprising:
training a machine learning model to determine distances between items of a plurality of items, wherein the training of the machine learning model comprises determining the distances based at least in part on assigning weights of a plurality of weights to a plurality of identifying markers of the plurality of items based at least in part on:
a first frequency by which identifying markers of the plurality of identifying markers appear in a particular item of the plurality of items;
a second frequency by which the identifying markers of the plurality of identifying markers appear in the plurality of items;
receiving, by a processor, a request that includes particular identifying markers that provide information about a query item that is being requested;
executing, using the processor, the machine learning model to determine a plurality of similarity distances between the query item and each particular item of a plurality of other items, wherein executing the machine learning model comprises:
determining a respective similarity distance for the particular item, of the plurality of similarity distances for the plurality of other items, based at least in part on respective weights of the plurality of weights and a cosine distance between the particular identifying markers associated with the query item and respective identifying markers of the particular item, wherein the respective similarity distance identifies a distance within a space associated with the machine learning model, wherein the plurality of other items are different than the query item; wherein the plurality of other items are filtered to include one or more other items with similarity distances within a predetermined range; wherein at least one other item outside the predetermined range is not included in the one or more other items;
accessing, by the processor and for each other item of the one or more other items, a respective entity profile comprising one or more scenario scores, wherein each of the one or more scenario scores is based on a metric associated with characterizing providing the other item by an entity associated with the respective entity profile; wherein the one or more scenario scores are specific to the entity and to the other item and do not include scenario scores specific to the entity and the at least one other item;
calculating, by the processor, and using the respective similarity distances and one or more scenario scores, an entity score for each respective entity profile to generate the entity list by ranking respective entities, associated with each respective entity profile, using the entity score; wherein calculating the entity score does not include a scenario score for the at least one other item even though the at least one other item has a scenario score for at least one entity in the entity list; and
outputting, by the processor, the entity list to a client device.
|