Abstract: Third party content providers can specify parameters for generating specialized search results in response to queries matching specific patterns. In this way, a generic search website can be enhanced to provide specialized search results to subscribed users. In one embodiment, these specialized results appear on a given user's result pages only when the user has subscribed to the enhancements from that particular content provider, so that users can tailor their search experience and see results that are more likely to be of interest to them. In other embodiments the specialized results are available to all users.

Abstract: A method and system for generating threads of documents from a collection C of documents containing terms. Each document of C has a timestamp and an associated timestamp index. The timestamp indexes are ordered in accordance with an ordering of the associated timestamps. A relevance graph G generated from C is an acyclic directed graph. Each node of G denotes a document of C. Each edge of G connects a pair of directed nodes pointing from a node having an earlier timestamp to a node having a later timestamp. At least one thread of G is determined by executing a matching-based algorithm or a dynamic programming algorithm. Each thread is a path through G originating at a first node and terminating at a second node and including one or more contiguous edges from the first node to the second node. The at least one thread is outputted.

Abstract: Context information, obtained by one programmable search engine and relating to a user, is stored in a central shared repository so that it can be used by other programmable search engines, in order to provide more effective targeted search results for the user in the future. Information obtained from other sources, including from the user him- or herself, can also be stored in the repository for future use. A programmable search engine consults the shared repository when processing queries and/or results, so as to incorporate previously stored context information concerning the user.

Abstract: Search results are generated using aggregated context data from two or more contexts. When two or more programmable search engines relate to a similar topic, context data associated with the programmable search engines are aggregated. The context is then applied to a query in order to present, in an integrated manner, relevant search results that make use of context intelligence from more than one programmable search engine.

Abstract: A programmable search engine system is programmable by a variety of different entities, such as client devices and vertical content sites to customize search results for users. Context files store instructions for controlling the operations of the programmable search engine. The contest files are processed by various context processors, which use the instructions therein to provide various pre-processing, post-processing, and search engine control operations.

Abstract: A programmable search engine system is programmable by a variety of different entities, such as client devices and vertical content sites to customize search results for users. Context files store instructions for controlling the operations of the programmable search engine. The context files are processed by various context processors, which use the instructions therein to provide various pre-processing, post-processing, and search engine control operations. Spam related and biased contexts and search results are identified using offline and query time processing stages, and the context files from vertical content providers associated with such spam and biased context and results are excluded from processing on direct user queries.

Abstract: Context, or user intent, is used for improving targeting of advertisements and for generating competition among advertisers for valuable ad space. Advertisers can bid for placement on search results pages based on combinations of keywords and context categories, or keywords and contexts. Such bids are compared to one another so that appropriate ads can be selected and displayed. By taking context into account, improved ad targeting is accomplished.

Abstract: A sales prediction system predicts sales from online public discussions. The system utilizes manually or automatically formulated predicates to capture subsets of postings in online public discussions. The system predicts spikes in sales rank based on online chatter. The system comprises automated algorithms that predict spikes in sales rank given a time series of counts of online discussions such as blog postings. The system utilizes a stateless model of customer behavior based on a series of states of excitation that are increasingly likely to lead to a purchase decision. The stateless model of customer behavior yields a predictor of sales rank spikes that is significantly more accurate than conventional techniques operating on sales rank data alone.

Abstract: A multi-structural query system performs a high-level multi-dimensional query on a multi-structural database. The query system enables a user to navigate a search by adding restrictions incrementally. The query system uses a schema to discover structure in a multi-structural database. The query system leaves a choice of nodes to return in response to a query as a constrained set of choices available to the algorithm. The query system further casts the selection of a set of nodes as an optimization. The query system uses pairwise-disjoint collections to capture a concise set of highlights of a data set within the allowed schema. The query system further comprises efficient algorithms that yield approximately optimal solutions for several classes of objective functions.

Abstract: A topic segmenting system segments a topic into chatter and subtopics. The system decomposes a conversation into topics, producing a time-based structure for topics and subtopics in the conversation. The system extracts a large number of topics at all levels of granularity. Some of the topics extracted correspond to broad topics and some correspond to “spiky” topics or subtopics. The system comprises a process for automatically detecting spiky regions of a topic. For each possible broad topic, the present system finds regions where coverage of the broad topic overlaps significantly with the spiky region of another topic. The system then removes the spiky subtopic from the conversation. Processing is repeated until all discernable topics have been identified and removed from the conversation, yielding random topics of little duration or intensity.

Abstract: A communication pattern inducing system focuses on the propagation of topics amongst a plurality of nodes based on the text of the node rather than hyperlinks of the node. A node could represent a weblog or any other source of information such as person, a conversation, images, etc. The system utilizes a model for information diffusion, wherein the parameters of the model capture how a new topic spreads from node to node. The system further comprises a process to learn the parameters of the model based on real data and to apply the process to real (or synthetic) node data. Consequently, the system is able to identify particular individuals that are highly effective at contributing to the spread of topics.