A method, system, and apparatus provide the ability to personalize a digital channel. A digital channel is provided to multiple users and visitor information at each visit is collected. The visitor information includes data about each visit and multiple content items that are presented. The users are autonomously clustered by segmenting the user population into behavioral groups such that mutual information is maximized between the users in an assigned behavioral group and the content items. Based on the clustering, a model is generated that estimates a score for each interaction between users and content items. The model is updated at a defined interval. Based on the score, content items to recommend to a specific user are determined. The recommendation jointly maximizes an outcome and a learning speed of the model. The personalized digital channel is delivered to the specific user based on the recommended multiple content items.

A system includes a set of crawlers that find and retrieve documents from an information network, an information extraction system, a knowledge graph storing nodes and edges that connect them, wherein each node represents a respective entity of a corresponding entity type of a plurality of entity types, and wherein the knowledge graph further stores event data relating to events detected by the information extraction system, a machine learning system that trains models that are used in connection with at least one of entity extraction, event extraction, recipient identification, and content generation, a lead scoring system that scores the relevance of information to an individual and references information in the knowledge graph, and a content generation system that generates content of a personalized message to a recipient who is an individual for which the lead scoring system has determined a threshold level of relevance.

Techniques for prefetching operation cost based digital content and digital content with emphasis that overcome the challenges of conventional systems are described. In one example, a computing device may receive digital content representations of digital content from a service provider system, which are displayed on a user interface of the computing device. Thereafter, the computing device may also receive digital content as prefetches having a changed display characteristic as emphasizing a portion of the digital content based on a model trained using machine learning. Alternatively, the computing device may receive digital content as a prefetch based on a model trained using machine learning in which the model addresses a likelihood of conversion of a good or service and an operation cost of providing the digital content. Upon receiving a user input selecting one of the digital content representations, digital content is rendered in the user interface of the computing device.

Method and system for localizing an element present in a piece of content having a plurality of elements. A cost of localizing an element with respect to each of one or more localization sources is first computed. At least one criterion based on which a localization source for localizing the element is to be determined is obtained. A localization source for to the element is then selected based on an assessment with respect to the at least one criterion. The element of the content is then localized using the selected localization source.

A computer-implemented method can facilitate delivery of targeted content to user devices in situations in which historic tracking data (e.g., cookie data) is generally unavailable and/or unreliable. A p-dimensional embedding of websites can be generated based on a group of user devices for whom tracking data is available. Conversion event data that indicates indicating whether that audience member performed a conversion action can be received. A machine learning model can be trained using the conversion event data and the positions of websites appearing in the conversion event data within the p-dimensional embedding to predict a likelihood of conversion and/or a type of content to provide given a position in the p-dimensional embedding. When an indication that a user device is accessing a website is received, a position of that website in the p-dimensional embedding can be determined and targeted content can be delivered to the user device.

A computer-implemented method can facilitate delivery of targeted content to user devices in situations in which historic tracking data (e.g., cookie data) is generally unavailable and/or unreliable. A p-dimensional embedding of websites can be generated based on a group of user devices for whom tracking data is available. Conversion event data that indicates indicating whether that audience member performed a conversion action can be received. A machine learning model can be trained using the conversion event data and the positions of websites appearing in the conversion event data within the p-dimensional embedding to predict a likelihood of conversion and/or a type of content to provide given a position in the p-dimensional embedding. When an indication that a user device is accessing a website is received, a position of that website in the p-dimensional embedding can be determined and targeted content can be delivered to the user device.

A system includes a network interface, a processing system, and a memory system. The memory system stores instructions that when executed by the processing system result in receiving a request and request data associated with a user from a web server and analyzing the request data to identify one or more data gaps associated with the request. One or more third-party services are called to fill at least a portion of the one or more data gaps. A question set is prepared based on determining that the one or more data gaps remain at least partially unfilled. The question set is selected by a machine-learning component trained to adapt a sequence and content of the question set over a plurality of interactions with a plurality of users. The question set is transmitted to the web server for presentation to the user. Data exchanges can be authenticated using tokens.

One or more computing devices, systems, and/or methods for isolated budget utilization are provided. A first budget pacing component is assigned to control bidding by a first content serving component for a set of content items. A second budget pacing component is assigned to control bidding by a second content serving component for the set of content items. The first budget pacing component controls the bidding by the first content serving component according to a first portion of a content item budget based upon a traffic share of the first content serving component. The second budget pacing component controls the bidding by the second content serving component according to a second portion of the content item budget based upon a traffic share of the second content serving component.

Methods, systems, and computer programs are presented for the determination of optimal communication scheduling. One method includes an operation for training a machine-learning program to generate a frequency model that determines a frequency for sending communications to users. The training utilizes training data defined by features related to user information and responses of users to previous communications to the users. The method further includes determining, by the frequency model and based on information about a first user, a first frequency for the first user. The first frequency identifies the number of communications to transmit to the first user per period of time. Further, the method includes operations for receiving a communication request to send one or more communications to the first user and determining send times for the one or more communications to the first user based on the first frequency. The communications are sent at the determined send times.

This paper presents a practical method for measuring the impact of multiple marketing events on sales, including marketing events that are not traditionally trackable. The technique infers which of several competing media events are likely to have caused a given conversion. The method is tested using hold-out sets, and also a live media experiment for determining whether the method can accurately predict television-generated web conversions.