A method, system and computer program product for ensuring a desired distribution of images in a multimedia document among different demographic groups. Demographic criteria (e.g., age) is received to form demographic groups of members of an organization. Demographic information along with interests of the members are retrieved. Such member data is analyzed within the constraints of the demographic criteria to generate a statistical distribution of members of the organization that forms an initially desired distribution of content of the multimedia document among the demographic groups of the organization. An indication is then provided to the user regarding whether the current distribution of the content of the multimedia document among the demographic groups of the organization satisfies or does not satisfy the desired distribution of content among the demographic groups. In this manner, the user can ensure a desired distribution of content in the multimedia document among different demographic groups.

An online system evaluates the quality of a content publisher displaying sponsored content items. To determine a likelihood of conversion actions associated with the sponsored content items, the online system uses information about users and their interactions with sponsored content items featured within the content publisher against interactions with sponsored content items featured within a benchmark system (e.g., online system). By determining a ratio of these interactions, the online system can determine a likelihood of conversion actions for the content publisher. The online system uses this likelihood of conversions to determine a publisher quality score that it uses to normalize third party value contributions toward placing sponsored content items on the content publisher. Thus, third party systems no longer need to be concerned about the intrinsic value of a given content publisher as third party value contributions are normalized based on the content publisher's conversion rates.

Dynamic selection of an advertisement to present to a user including detecting, by an advertisement selection module, a proximity between the user and a device including retrieving sensor data from at least one sensor of the device; generating, by the advertisement selection module, a current environmental profile comprising the detected proximity between the user and a device; matching, by the advertisement selection module, the current environmental profile to a first personal skip profile of a plurality of personal skip profiles for the user, wherein the first personal skip profile comprises data indicating a likelihood that the user will skip the advertisement for the detected proximity; selecting the advertisement based on the first personal skip profile including the data indicating the likelihood that the user will skip the advertisement for the detected proximity; and sending, to the device, the selected advertisement for presentation to the user.

Techniques are provided for k-best option identification of options subject to a supplied tolerance. One technique includes: sampling the options for a first period on a plurality of computers; computing an average and a sample count for each option based on the sampling; splitting the options into a highest group and a lowest group based on the computed averages; selecting a weakest one of the highest group (option A) and a strongest one of the lowest group (option B); and deciding whether or not to terminate based on the supplied tolerance and the selecting of options A and B. In some cases, the technique further includes outputting the highest group and terminating in response to a termination decision; otherwise continue with sampling options A and B for a next period; and updating the computed average and the sample count for options A and B based on corresponding next period sampling.

Influence maximization determination within a social network system is described. In one example, a subset is selected from a plurality of user accounts of a social network system. Exposure of digital marketing content is then caused to the subset of user accounts. A determination is made as to a probability of each user account of the plurality of user accounts as being influenced by the exposure of the digital marketing content to the subset of user accounts. The determined probability is then output, such as to control output of digital marketing content.

Rules for triggering geo-notifications, such as push notifications to a mobile computing device based on its location, are automatically generated based on a user's location, dwell time, and browsing activity. In some embodiments, a geo-fence communication system determines an average dwell time for a group of users in a user segment. The average dwell time indicates an average time for each user to respond to a geo-notification for a product, within a geo-fence. The geo-fence communication system generates a rule for the user segment/product/geo-fence tuple based on the average dwell time. In some cases, semantic similarities between geo-fence locations are determined, and a rule for a geo-fence is generated based on average dwell times for other geo-fences in similar locations.

Targeting information (also referred to as ad serving constraints) or candidate targeting information for an advertisement is identified. Targeting information may be identified by extracting topics or concepts from, and/or generating topics or concepts based on, ad information, such as information from a Web page to which an ad is linked (or some other Web page of interest to the ad or advertiser). The topics or concepts may be relevant queries associated with the Web page of interest, clusters, etc.

A system and method are disclosed for selecting advertisements based on local intent. Local intent may reflect whether a search query should receive results and advertisements that are geographically specific. The local intent may be determined using probabilistic models that analyze historical searches to determine which search terms tend to have local intent.

A user behavior monitoring module logs the operation of an image browser and sends the logged information to a personalization system. The personalization system includes a monitoring management module, a client statistics data store, a recommender system, and a content refinement module. The monitoring management module receives the logged information sent by the user behavior monitoring module and stores it in the client statistics data store. The recommender system determines user preferences based on the information stored in the client statistics data store. The content refinement module refines a set of content items based on the user preferences determined by the recommender system. The set is refined so that the resulting content items (and their ordering) are more relevant to the user operating the image browser. The refined set of content items is displayed to the user by the image browser.

A system for alerting an employee or agent of a retailer regarding an unfavorable condition may include the use of computer-aided visual recognition of products to aid in identifying the location of the unfavorable condition. A user can direct a mobile computing device camera at one or more targeted products displayed on store shelves. Image recognition operations can be carried out to compare the targeted product image(s) against images from a prepopulated product image store of known products. Upon a positive match, shelf location information of the identified targeted product(s) may be used to determine the current location of the user. The employee or agent may be alerted with respect to the unfavorable condition and the location thereof.