Determining intent based on user interaction data

Abstract

A system predicts user intent to take an action and delivers content items to the user that match that intent. A plurality of features or attributes for each tracking pixel in a set of tracking pixels can be acquired based on content items and landing pages associated with each tracking pixel. For example, features for a tracking pixel can be determined based on information associated with a content item that enabled a user to access a landing page from which the tracking pixel was fired or triggered. In this example, features for the tracking pixel can also be determined based on information associated with the landing page. The features for the tracking pixels can be utilized to train a machine learning model. The machine learning model can be trained to predict whether or not a particular user intends to produce a conversion (e.g., make a purchase).

What is claimed is: 
     
         1 . A computer-implemented method comprising:
 generating, by an online system, a plurality of features for each tracking pixel in a set of tracking pixels used by a third party system to track actions of users on the third party system, wherein each tracking pixel is associated with a particular action that a given content item and given landing page are directed to getting a user to take;   generating, by the online system, a plurality of features for the user based on interaction data describing prior interactions by the user related to content provided to the user that ultimately led to the user taking a previous particular action;   training, by the online system, a machine learning model based on the plurality of features for each tracking pixel to predict an intent of the user to take a future particular action based on the user having had interactions similar to those that ultimately led to the user taking the previous particular action;   applying, by the online system, the machine learning model to predict the intent of the user to take the future particular action and to select content items directed to getting the user to take the future particular action; and   providing, by the online system, a particular content item of the selected content items to the user based on the intent predicted for the user.   
     
     
         2 . The computer-implemented method of  claim 1 , further comprising:
 providing, for review, information associated with a particular tracking pixel in the set of tracking pixels;   acquiring feedback from the review of the information associated with the particular tracking pixel; and   applying, based on the feedback from the review, one or more weights to one or more features for the particular tracking pixel, wherein the machine learning model is trained based on the one or more features subsequent to the one or more weights being applied.   
     
     
         3 . The computer-implemented method of  claim 2 , wherein the review of the information associated with the particular tracking pixel is based on manual effort. 
     
     
         4 . The computer-implemented method of  claim 1 , wherein the given content item and the given landing page associated with each tracking pixel are reviewed based on manual effort prior to generating the plurality of features for each tracking pixel in the set of tracking pixels. 
     
     
         5 . The computer-implemented method of  claim 1 , further comprising:
 determining, based on the machine learning model, another confidence score representing another likelihood that another conversion will result from the user being provided with another content item via the online system; and   selecting, based on the confidence score and the other confidence score, respectively, to present to the user at least one of the particular content item or the other content item.   
     
     
         6 . The computer-implemented method of  claim 1 , further comprising:
 determining a level of specificity associated with the intent of the user to take the future particular action.   
     
     
         7 . The computer-implemented method of  claim 6 , wherein the level of specificity indicates that the intent of the user to take the future particular action is associated with at least one of a particular product, a particular brand, a particular manufacturer, a particular retailer, or a particular product type. 
     
     
         8 . The computer-implemented method of  claim 1 , wherein at least one landing page in the set of one or more landing pages is different from the online system. 
     
     
         9 . The computer-implemented method of  claim 1 , wherein the interaction data associated with the user includes information provided by at least one tracking pixel associated with the user, wherein the at least one tracking pixel is implemented via at least one landing page, and wherein the information provided by the at least one tracking pixel includes at least one of information about an event at the at least one landing page, information about a time of a user interaction with respect to the at least one landing page, information about a duration of the user interaction with respect to the at least one landing page, information about one or more properties of the at least one landing page, information about one or more properties of at least one content item that enabled the user to access the at least one landing page, or information about the user. 
     
     
         10 . A system comprising:
 at least one processor; and   a memory storing instructions that, when executed by the at least one processor, cause the system to perform:
 generating a plurality of features for each tracking pixel in a set of tracking pixels used by a third party system to track actions of users on the third party system, wherein each tracking pixel is associated with a particular action that a given content item and given landing page are directed to getting a user to take; 
 generating a plurality of features for the user based on interaction data describing prior interactions by the user related to content provided to the user that ultimately led to the user taking a previous particular action; 
 training a machine learning model based on the plurality of features for each tracking pixel generated for the user to predict an intent of the user to take a future particular action based on the user having had interactions similar to those that ultimately led to the user taking the previous particular action; 
 applying the machine learning model to predict the intent of the user to take the future particular action and to select content items directed to getting the user to take the future particular action; and 
 providing a particular content item of the selected content items to the user based on the intent predicted for the user. 
   
     
     
         11 . The system of  claim 10 , wherein the instructions cause the system to further perform:
 providing, for review, information associated with a particular tracking pixel in the set of tracking pixels;   acquiring feedback from the review of the information associated with the particular tracking pixel; and   applying, based on the feedback from the review, one or more weights to one or more features for the particular tracking pixel, wherein the machine learning model is trained based on the one or more features subsequent to the one or more weights being applied.   
     
     
         12 . The system of  claim 11 , wherein the review of the information associated with the particular tracking pixel is based on manual effort. 
     
     
         13 . The system of  claim 10 , wherein the given content item and the given landing page associated with each tracking pixel are reviewed based on manual effort prior to generating the plurality of features for each tracking pixel in the set of tracking pixels. 
     
     
         14 . The system of  claim 10 , wherein the instructions cause the system to further perform:
 determining a level of specificity associated with the intent of the user to take the future particular action.   
     
     
         15 . The system of  claim 14 , wherein the level of specificity indicates that the intent of the user to take the future particular action is associated with at least one of a particular product, a particular brand, a particular manufacturer, a particular retailer, or a particular product type. 
     
     
         16 . A non-transitory computer-readable storage medium comprising instructions that when executed cause a processor to:
 generate, by an online system, a plurality of features for each tracking pixel in a set of tracking pixels used by a third party system to track actions of users on the third party system, wherein each tracking pixel is associated with a particular action that a given content item and given landing page are directed to getting a user to take;   generate, by the online system, a plurality of features for the user based on interaction data describing prior interactions by the user related to content provided to the user that ultimately led to the user taking a previous particular action;   train, by the online system, a machine learning model based on the plurality of features for each tracking pixel to predict an intent of the user to take a future particular action based on the user having had interactions similar to those that ultimately led to the user taking the previous particular action;   determine a level of specificity associated with the intent of the user to take the future particular action;   apply, by the online system, the machine learning model to predict the intent of the user to take the future particular action and to select content items directed to getting the user to take the future particular action based at least in part on the determined level of specificity; and   provide, by the online system, a particular content item of the selected content items to the user based on the intent predicted for the user.   
     
     
         17 . The non-transitory computer-readable storage medium of  claim 16 , further comprising:
 providing, for review, information associated with a particular tracking pixel in the set of tracking pixels;   acquiring feedback from the review of the information associated with the particular tracking pixel; and   applying, based on the feedback from the review, one or more weights to one or more features for the particular tracking pixel, wherein the machine learning model is trained based on the one or more features subsequent to the one or more weights being applied.   
     
     
         18 . The non-transitory computer-readable storage medium of  claim 16 , wherein the given content item and the given landing page associated with each tracking pixel are reviewed based on manual effort prior to generating the plurality of features for each tracking pixel in the set of tracking pixels. 
     
     
         19 . The non-transitory computer-readable storage medium of  claim 16 , wherein the intent of the user to take the future particular action is predicted based on determining a confidence score representing a likelihood that a conversion will result from the user being provided with the particular content item via the online system. 
     
     
         20 . The non-transitory computer-readable storage medium of  claim 16 , wherein the level of specificity indicates that the intent of the user to take the future particular action is associated with at least one of a particular product, a particular brand, a particular manufacturer, a particular retailer, or a particular product type.