Abstract: A real-time messaging platform allows advertiser accounts to pay to insert candidate messages into the message streams requested by account holders. To accommodate multiple advertisers, the messaging platform controls an auction process that determines which candidate messages are selected for inclusion in a requested account holder's message stream. Selection is based on a bid for the candidate message, the message stream that is requested, and a variety of other factors that vary depending upon the implementation. The process for selection of candidate messages generally includes the following steps, though any given step may be omitted or combined into another step in a different implementation: targeting, filtering, prediction, ranking, and selection.

Abstract: A real-time messaging platform allows advertiser accounts to pay to insert candidate messages into the message streams requested by account holders. To accommodate multiple advertisers, the messaging platform controls an auction process that determines which candidate messages are selected for inclusion in a requested account holder's message stream. Selection is based on a bid for the candidate message, the message stream that is requested, and a variety of other factors that vary depending upon the implementation. The process for selection of candidate messages generally includes the following steps, though any given step may be omitted or combined into another step in a different implementation: targeting, filtering, prediction, ranking, and selection.

Abstract: A real-time messaging platform allows advertiser accounts to pay to insert candidate messages into the message streams requested by account holders. To accommodate multiple advertisers, the messaging platform controls an auction process that determines which candidate messages are selected for inclusion in a requested account holder's message stream. Selection is based on a bid for the candidate message, the message stream that is requested, and a variety of other factors that vary depending upon the implementation. The process for selection of candidate messages generally includes the following steps, though any given step may be omitted or combined into another step in a different implementation: targeting, filtering, prediction, ranking, and selection.

Abstract: A real-time messaging platform allows advertiser accounts to pay to insert candidate messages into the message streams requested by account holders. To accommodate multiple advertisers, the messaging platform controls an auction process that determines which candidate messages are selected for inclusion in a requested account holder's message stream. Selection is based on a bid for the candidate message, the message stream that is requested, and a variety of other factors that vary depending upon the implementation. The process for selection of candidate messages generally includes the following steps, though any given step may be omitted or combined into another step in a different implementation: targeting, filtering, prediction, ranking, and selection.

Abstract: A real-time messaging platform allows advertiser accounts to pay to insert candidate messages into the message streams requested by account holders. To accommodate multiple advertisers, the messaging platform controls an auction process that determines which candidate messages are selected for inclusion in a requested account holder's message stream. Selection is based on a bid for the candidate message, the message stream that is requested, and a variety of other factors that vary depending upon the implementation. The process for selection of candidate messages generally includes the following steps, though any given step may be omitted or combined into another step in a different implementation: targeting, filtering, prediction, ranking, and selection.

Abstract: Methods, systems, and computer-storage media having computer-usable instructions embodied thereon for calculating event probabilities are provided. The event may be a click probability. Event probabilities are calculated using a system optimized for runtime model accuracy with an operable learning algorithm. Bin counting techniques are used to calculate event probabilities based on a count of event occurrences and non-event occurrences. Linear parameters, such and counts of clicks and non-clicks, may also be used in the system to allow for runtime adjustments.

Abstract: A client application installed on end user computers generates metadata from the content of web pages visited by end users and provides the metadata to a search engine. When an end user visits a web page, the end user's computer downloads and displays the web page to the end user. The client application may simultaneously access the web page content and generate this metadata in the form of a content signature of the web page from the web page content. The client application then provides the content signature to a search engine. The search engine may employ content signatures to identify new web pages to crawl and index. Additionally, the search engine may employ content signatures to identify changes to web pages and determine the crawl frequency of web pages.

Abstract: A human or hand-labeled ranking of URL results for a search query is compared against actual click data for the respective query/URL pairs (e.g., which URLs were actually clicked on by users when the URLs were presented to users when the search query was run in the real world). The human ranking or ordering of the URL results (e.g., pre-existing relevance ranking) for the query can then be adjusted, if necessary, based upon the real world click data (e.g., click relevance ranking). The modified pre-existing relevance ranking can be used in providing future search results.

Abstract: A human or hand-labeled ranking of URL results for a search query is compared against actual click data for the respective query/URL pairs (e.g., which URLs were actually clicked on by users when the URLs were presented to users when the search query was run in the real world). The human ranking or ordering of the URL results (e.g., pre-existing relevance ranking) for the query can then be adjusted, if necessary, based upon the real world click data (e.g., click relevance ranking). The modified pre-existing relevance ranking can be used in providing future search results.

Abstract: Methods, systems, and computer-storage media having computer-usable instructions embodied thereon for calculating event probabilities are provided. The event may be a click probability. Event probabilities are calculated using a system optimized for runtime model accuracy with an operable learning algorithm. Bin counting techniques are used to calculate event probabilities based on a count of event occurrences and non-event occurrences. Linear parameters, such and counts of clicks and non-clicks, may also be used in the system to allow for runtime adjustments.

Abstract: Short paths are found with a small query time in scale-free directed graphs using a two-phase process by which data structures comprising shortest path trees are first pre-computed for a group of central vertices called “hubs” that have short paths to most other vertices in the graph. In a query time phase, a short path between two vertices of interest in the graph is found by looking up the path to the root in each of the shortest path trees.

Abstract: A human or hand-labeled ranking of URL results for a search query is compared against actual click data for the respective query/URL pairs (e.g., which URLs were actually clicked on by users when the URLs were presented to users when the search query was run in the real world). The human ranking or ordering of the URL results (e.g., pre-existing relevance ranking) for the query can then be adjusted, if necessary, based upon the real world click data (e.g., click relevance ranking). The modified pre-existing relevance ranking can be used in providing future search results.

Abstract: A method for classifying search query traffic can involve receiving a plurality of labeled sample search query traffic and generating a feature set partitioned into human physical limit features and query stream behavioral features. A model can be generated using the plurality of labeled sample search query traffic and the feature set. Search query traffic can be received and the model can be utilized to classify the received search query traffic as generated by a human or automatically generated.

Abstract: Classes of web graph algorithms are extended to run directly on virtual node-type compressed web graphs where a reduction in runtime of the extended algorithms is realized which is approximately proportional to the compression ratio applied to the original (i.e., uncompressed) graph. In the virtual node compression technique, a succinct representation of a web graph is constructed by replacing dense subgraphs by sparse ones so that the resulting compressed graph has significantly fewer edges and a relatively small number of additional nodes.

Abstract: A client application installed on end user computers generates metadata from the content of web pages visited by end users and provides the metadata to a search engine. When an end user visits a web page, the end user's computer downloads and displays the web page to the end user. The client application may simultaneously access the web page content and generate this metadata in the form of a content signature of the web page from the web page content. The client application then provides the content signature to a search engine. The search engine may employ content signatures to identify new web pages to crawl and index. Additionally, the search engine may employ content signatures to identify changes to web pages and determine the crawl frequency of web pages.

Abstract: A human or hand-labeled ranking of URL results for a search query is compared against actual click data for the respective query/URL pairs (e.g., which URLs were actually clicked on by users when the URLs were presented to users when the search query was run in the real world). The human ranking or ordering of the URL results (e.g., pre-existing relevance ranking) for the query can then be adjusted, if necessary, based upon the real world click data (e.g., click relevance ranking). The modified pre-existing relevance ranking can be used in providing future search results.

Abstract: Classes of web graph algorithms are extended to run directly on virtual node-type compressed web graphs where a reduction in runtime of the extended algorithms is realized which is approximately proportional to the compression ratio applied to the original (i.e., uncompressed) graph. In the virtual node compression technique, a succinct representation of a web graph is constructed by replacing dense subgraphs by sparse ones so that the resulting compressed graph has significantly fewer edges and a relatively small number of additional nodes.

Abstract: Short paths are found with a small query time in scale-free directed graphs using a two-phase process by which data structures comprising shortest path trees are first pre-computed for a group of central vertices called “hubs” that have short paths to most other vertices in the graph. In a query time phase, a short path between two vertices of interest in the graph is found by looking up the path to the root in each of the shortest path trees.

Abstract: Multiple Bloom filters are generated to partition data between first and second disjoint data sets of elements. Each element in the first data set is assigned to a bucket of a first set of buckets, and each element in the second data set is assigned to a bucket of a second set of buckets. A Bloom filter is generated for each bucket of the first set of buckets. The Bloom filter generated for a bucket indicates that each element assigned to that bucket is part of the first data set, and that each element assigned to a corresponding bucket of the second set of buckets is not part of the first data set. Additionally, a Bloom filter corresponding to a subsequently received element can be determined and used to identify whether that subsequently received element is part of the first data set or the second data set.

Abstract: A method for classifying search query traffic can involve receiving a plurality of labeled sample search query traffic and generating a feature set partitioned into human physical limit features and query stream behavioral features. A model can be generated using the plurality of labeled sample search query traffic and the feature set. Search query traffic can be received and the model can be utilized to classify the received search query traffic as generated by a human or automatically generated.