Abstract: Search results responsive to a query are generated by a search engine. If the query is “location relevant”, the results are provided to a ranker component. If the query is an entity search, then entities matching the query are ranked by distance from the user. If the query is not, and if it is not semantically charged, the entities are ranked either according to distance from the user or by search relevance. If the query is semantically charged, a semantic score is calculated for each entity. One or more threshold semantic scores are determined that divide the search results into multiple parts. The results within each part can then be ranked independently. The results within a first part are ranked according to their distance from the user, while the results in a subsequent part are ranked according to their semantical relationship to the query. The re-ranked parts are concatenated together.

Abstract: An apparatus for a security system is described. The apparatus includes a plurality of security cameras, a monitoring processor that tracks a focus of attention of a user via selection of cameras one at a time from among the plurality of security cameras, a collection processor that detects an evidence collection request from the user, opens an evidence file in response thereto and saves a sequence of video images from at least some of the plurality of video cameras into the evidence file based upon the tracked focus of attention of the user.

Abstract: Search results responsive to a query are generated by a search engine. If the query is “location relevant”, the results are provided to a ranker component. If the query is an entity search, then entities matching the query are ranked by distance from the user. If the query is not, and if it is not semantically charged, the entities are ranked either according to distance from the user or by search relevance. If the query is semantically charged, a semantic score is calculated for each entity. One or more threshold semantic scores are determined that divide the search results into multiple parts. The results within each part can then be ranked independently. The results within a first part are ranked according to their distance from the user, while the results in a subsequent part are ranked according to their semantical relationship to the query. The re-ranked parts are concatenated together.

Abstract: Ranking of documents by metadata popularity provides relevant search results in response to user search queries received by a search engine. Metadata popularity is determined by comparing metadata from a document with popularity data from one or more sources. In some embodiments, metadata popularity is determined based on a frequency with which extracted metadata appears in query logs. Search results are ordered based on metadata popularity and returned in response to the user search queries.

Abstract: Embodiments of the present invention relate to aggregating opinion data and providing aggregated opinion data to a user. More specifically, opinion data may be aggregated by one or more categories, or by entity. Opinion data may be received from various sources and includes sentiment information (e.g., positive and negative comments) on a specific entity. The opinion data may be further aggregated by feature for each category.