Abstract: Systems and methods for recommending bidded terms are disclosed. The system collects a plurality of bidded terms and separates them into ad groups. The add groups are then combined into sequences of terms, which are fed into a deep learning network to build a multidimensional word vector in which related terms are nearer one another than unrelated terms. An input term is then received and the system matches the input term in the multidimensional word vector and recommends the nearest neighbors to the term.

Abstract: Embodiments disclosed herein may relate to identifying music from a text source utilizing a computing platform in a communication system.

Abstract: A computer-implemented method serves content to a device. The method includes identifying a characteristic of content displayed with a first device. Through the use of a processor, the characteristic is compared with a campaign constraint of an advertising campaign that includes an advertisement. The method includes transmitting the advertisement to a second device for display when the characteristic satisfies the campaign constraint.

Abstract: The present invention provided techniques that may be used, for example, in online advertising. Techniques are provided that include centralized and aggregated advertisement performance data tracking, which can be used in advertisement selection. Advertisement performance data may be obtained by a central server, such as from many advertisement servers. The central server may aggregate the performance data, and may generate a performance snapshot spanning many advertisements. The snapshot may be used in generating a performance prediction model, which may in turn be used in advertisement selection, or monitoring or tracking associated with advertising. Elements or actions, such as obtaining of performance data, aggregation, generation of the snapshot, and generation of the model, may include frequent or real time updating of such elements or actions.

Abstract: Techniques are provided that utilize online advertising traffic patterns at multiple geographically distributed data centers to make predictions about future traffic at the data centers, and use the predicted patterns in optimizing proportioning or allocation of an advertising budget between each of the data centers. For each of multiple geographically distributed data centers, a set of traffic data may be obtained, relating to past online advertising traffic at the data center. For each of the data centers, the set of traffic data is used in determining a prediction of online advertising traffic at the data center of a future period of time, which may include determining and utilizing a traffic pattern function for the data center. Using the prediction for each of the data centers, an optimized allocation is performed of an advertiser's online advertising budget over a future period of time between each of the data centers.

Abstract: A method, system and computer readable medium for recommending new connections in an advertising exchange. The method identifies participants that share some similarity (e.g. a pre-existing advertising relationship), and on the basis of the features and extent of a similarity, and then suggests or recommends other relationships that can also be shared. The method commences by enumerating a set of producer participants and by enumerating a set of consumer participants where at least two of the members of the set of consumer participants has an ad relationship to at least one of the producer participants. Then, on the basis of a selected feature quantity the extent of the similarity is calculated. When it is found that one consumer is similar to a second consumer (by virtue of at least one shared ad relationship), then the system recommends to the participants one or more additional ad relationships that can be established.

Abstract: Methods and apparatus are described to automatically categorize search queries. According to specific embodiments, this is accomplished by comparing search results responsive to an uncategorized query with search results responsive to queries in a categorized set. Search results from the categorized set are assigned categories and weights corresponding to the queries which produced them. Matches with search results for the uncategorized query are located in this data, and the corresponding categories and weights associated with the uncategorized query. These techniques can be applied to improve the relevancy of organic search results, sponsored search results, advertisements, marketing communications, news articles, and other types of content on both the provider's websites and other websites.

Abstract: A computer based method of biometric analysis which compares a first vector from a first biometric sample with a second vector from a second biometric sample. The vectors have at least one biometric feature. A method which compares two biometric samples. The samples form at least one cluster of at least one vector based on feature similarities between the samples. An apparatus incorporating a means for performing the method taught herein. A computer-readable medium and a propagated computer data signal transmitted via a propagation medium, with instructions which when executed by a processor, carry out the method of the present invention. A method of handwriting analysis which calculates a first metric from a first vector having at least one feature from a sample, calculates a second metric from a second vector having at least one feature from a second sample, and calculates the distance in two-dimensional feature space between metrics.