Ranked relevance results using multi-feature scoring returned from a universal relevance service framework

Abstract

In general, embodiments of the present invention provide systems, methods and computer readable media for a universal relevance service framework for ranking and personalizing items.

1 - 20 . (canceled) 
     
     
         21 . An apparatus comprising at least one processor and at least one non-transitory computer-readable storage medium storing instructions that, when executed by the at least one processor, cause the apparatus to:
 generate one or more trained machine learning models based at least in part on supervised learning models and training data, wherein the one or more trained machine learning models are configured for generating a user-deal relevance score based at least in part on one or more feature vectors associated with user-deal interaction data associated with a particular user;   receive, from a relevance API client device, a search request associated with the particular user, the search request comprising a plurality of search query terms;   retrieve, from a data store, a list of feature data associated with a plurality of products identified based on the plurality of search query terms;   generate, based on a set of join tables retrieved from the data store, a plurality of feature vectors, wherein the set of join tables comprises user-interaction data describing user-deal interactions for the particular user;   generate a user-deal relevance score for each product of the plurality of products based at least in part on applying the one or more trained machine learning models to the plurality of feature vectors and the list of feature data;   rank the plurality of products according to their respective user-deal relevance scores; and   transmit, to the relevance API client device, a top N ranked products of the plurality of products for configured for display within the electronic interface of a user device.   
     
     
         22 . The apparatus of  claim 21 , wherein the list of feature data comprises past performance data computed from jointed user and deal attributes, individual product performance data associated with all users similar to the particular user, a computation of the likelihood that a preference of the particular user is similar to the product, a first penalty based on whether a product has been exposed previously to the particular user, or a second penalty if the particular user has purchased the product recently. 
     
     
         23 . The apparatus of  claim 21 , wherein the set of join tables comprise one or more of a user-deal (U×D) join table, a user-deal-attribute (U×DA) join table, a user-attribute-deal (UA×D) join table, or a user-attribute-deal-attribute (UA×DA) join table. 
     
     
         24 . The apparatus of  claim 23 , wherein the set of join tables stores data comprising one or more of send events, impression events, view events, click events, purchase events, counts, raw counts, or decayed counts. 
     
     
         25 . The apparatus of  claim 24 , wherein generating the plurality of feature vectors comprises transforming the join tables into matrices and constructing the plurality of feature vectors based on the matrices. 
     
     
         26 . The apparatus of  claim 21 , wherein a first feature of a feature vector is weighted more heavily than a second feature based on a determination that the first feature is more important relative to the second feature for separating classes. 
     
     
         27 . The apparatus of  claim 21 , wherein the at least one non-transitory computer-readable storage medium stores instructions that, when executed by the at least one processor, further cause the apparatus to:
 determine that a ratio of a click likelihood for the particular user to an average click likelihood exceeds a predefined number;   adjust the user-deal relevance score based at least in part on the ratio to generate an adjusted user-deal relevance score;   generate a re-ranked list of the plurality of products based on the adjusted user-deal relevance score; and   transmit, to the relevance API client device, a top N ranked products of re-ranked list of the plurality of products for configured for display within the electronic interface of a user device.   
     
     
         28 . A computer-implemented method, comprising:
 generating, using a processor, one or more trained machine learning models based at least in part on supervised learning models and training data, wherein the one or more trained machine learning models are configured for generating a user-deal relevance score based at least in part on one or more feature vectors associated with user-deal interaction data associated with a particular user;   receiving, using the processor and from a relevance API client device, a search request associated with the particular user, the search request comprising a plurality of search query terms;   retrieving, using the processor and from a data store, a list of feature data associated with a plurality of products identified based on the plurality of search query terms;   generating, using the processor and based on a set of join tables retrieved from the data store, a plurality of feature vectors, wherein the set of join tables comprises user-interaction data describing user-deal interactions for the particular user;   generating, using the processor, a user-deal relevance score for each product of the plurality of products based at least in part on applying the one or more trained machine learning models to the plurality of feature vectors and the list of feature data;   ranking, using the processor, the plurality of products according to their respective user-deal relevance scores; and   transmitting, using the processor and to the relevance API client device, a top N ranked products of the plurality of products for configured for display within the electronic interface of a user device.   
     
     
         29 . The method of  claim 28 , wherein the list of feature data comprises past performance data computed from jointed user and deal attributes, individual product performance data associated with all users similar to the particular user, a computation of the likelihood that a preference of the particular user is similar to the product, a first penalty based on whether a product has been exposed previously to the particular user, or a second penalty if the particular user has purchased the product recently. 
     
     
         30 . The method of  claim 28 , wherein the set of join tables comprise one or more of a user-deal (U×D) join table, a user-deal-attribute (U×DA) join table, a user-attribute-deal (UA×D) join table, or a user-attribute-deal-attribute (UA×DA) join table. 
     
     
         32 . The method of claim  31 , wherein the set of join tables stores data comprising one or more of send events, impression events, view events, click events, purchase events, counts, raw counts, or decayed counts. 
     
     
         33 . The method of  claim 32 , wherein generating the plurality of feature vectors comprises transforming the join tables into matrices and constructing the plurality of feature vectors based on the matrices. 
     
     
         34 . The method of  claim 28 , wherein a first feature of a feature vector is weighted more heavily than a second feature based on a determination that the first feature is more important relative to the second feature for separating classes. 
     
     
         35 . The method of  claim 28 , further comprising:
 determining, using the processor, that a ratio of a click likelihood for the particular user to an average click likelihood exceeds a predefined number;   adjusting, using the processor, the user-deal relevance score based at least in part on the ratio to generate an adjusted user-deal relevance score;   generating, using the processor, a re-ranked list of the plurality of products based on the adjusted user-deal relevance score; and   transmitting, using the processor and to the relevance API client device, a top N ranked products of re-ranked list of the plurality of products for configured for display within the electronic interface of a user device.   
     
     
         36 . A non-transitory computer readable medium storing instructions that, when executed by at least one processor of an apparatus, cause the apparatus to:
 generate one or more trained machine learning models based at least in part on supervised learning models and training data, wherein the one or more trained machine learning models are configured for generating a user-deal relevance score based at least in part on one or more feature vectors associated with user-deal interaction data associated with a particular user;   receive, from a relevance API client device, a search request associated with the particular user, the search request comprising a plurality of search query terms;   retrieve, from a data store, a list of feature data associated with a plurality of products identified based on the plurality of search query terms;   generate, based on a set of join tables retrieved from the data store, a plurality of feature vectors, wherein the set of join tables comprises user-interaction data describing user-deal interactions for the particular user;   generate a user-deal relevance score for each product of the plurality of products based at least in part on applying the one or more trained machine learning models to the plurality of feature vectors and the list of feature data;   rank the plurality of products according to their respective user-deal relevance scores; and   transmit, to the relevance API client device, a top N ranked products of the plurality of products for configured for display within the electronic interface of a user device.   
     
     
         37 . The computer readable medium of  claim 36 , wherein the list of feature data comprises past performance data computed from jointed user and deal attributes, individual product performance data associated with all users similar to the particular user, a computation of the likelihood that a preference of the particular user is similar to the product, a first penalty based on whether a product has been exposed previously to the particular user, or a second penalty if the particular user has purchased the product recently. 
     
     
         38 . The computer readable medium of  claim 36 , wherein the set of join tables comprise one or more of a user-deal (U×D) join table, a user-deal-attribute (U×DA) join table, a user-attribute-deal (UA×D) join table, or a user-attribute-deal-attribute (UA×DA) join table. 
     
     
         39 . The computer readable medium of  claim 38 , wherein the set of join tables stores data comprising one or more of send events, impression events, view events, click events, purchase events, counts, raw counts, or decayed counts. 
     
     
         40 . The computer readable medium of  claim 39 , wherein generating the plurality of feature vectors comprises transforming the join tables into matrices and constructing the plurality of feature vectors based on the matrices.