초록

An electronic document sorter is trained to classify documents based on their temporal qualities. The invention can be used in environments such as automated news aggregators, search engines and other electronic systems which compile information having temporal qualities.

대표청구항 ▼

1. A method of automatically training an electronic document sorter with a computing system comprising: a) identifying a first set of electronic documents containing content associated with a set of topics using the computing system;wherein said content describes one or more events related to one or

1. A method of automatically training an electronic document sorter with a computing system comprising: a) identifying a first set of electronic documents containing content associated with a set of topics using the computing system;wherein said content describes one or more events related to one or more topics;b) analyzing said first set of electronic documents to identify first temporal components for each of said one or more events and one or more topics;wherein said first temporal components include content other than but not excluding date/time data specified for said first set of electronic documents, including text words and semantic equivalents of such text words;c) assigning a first set of corresponding temporal scores to said first set of electronic documents based on analyzing said first temporal components, said first set of temporal scores representing a first set of chronological states of said one or more events for said one or more topics described in said first set of documents,wherein said first set of corresponding temporal scores are provided by automated and/or human based input scoring;d) training an automated document classifier using said temporal components; wherein the automated document classifier is adapted by such training to classify and assign a second temporal score to an unclassified electronic document containing content associated with the first event related to a first topic based on temporal characteristics analyzed for such unclassified electronic document, including second content other than but not excluding second date/time data including second text words and second semantic equivalents thereof representing a second set of temporal components contained in such unclassified document;wherein said unclassified document is ordered in time relative to said first set of documents with respect to at least said first event based on said second temporal score. 2. The method of claim 1 wherein said second temporal score indicates a specified action in a sequence of actions taken by a subject of said first event. 3. The method of claim 1 wherein said automated document classifier is further trained to determine an event associated with an unclassified electronic document, and receives temporal component information from a reference document set containing reference temporal data corresponding to predicted content for said event. 4. The method of claim 3 wherein said predicted content includes one or more objects and related object states for said predicted content. 5. The method of claim 1 wherein said first temporal components are derived from annotations and/or codings made to such first set of electronic documents provided by a human reviewer. 6. The method of claim 1 wherein said first temporal components are derived from annotations and/or codings made automatically to said first set of electronic documents by a text parser. 7. The method of claim 1 wherein said automated document classifier is adapted to use semantic data for a sport event content associated with said unclassified electronic document. 8. The method of claim 1 further including a step tuning said automated document classifier with additional temporal related data for a class of events derived from a subsequent set of electronic documents relating to such class of events. 9. The method of claim 1 wherein said events are sports related, natural disaster related, accident related, political election related, and/or financial market related. 10. The method of claim 1 wherein said automated document classifier is used with an automated news aggregator executing on a web site. 11. The method of claim 1 wherein said automated document classifier is used with a search engine. 12. The method of claim 1 wherein said automated document classifier is used with a recommender system. 13. The method of claim 1 wherein said temporal components are stored as selected content depicting separate sequential temporal states of said event. 14. The method of claim 13, wherein said selected content also includes variable data designed to accommodate data which indirectly indicates a temporal state. 15. The method of claim 14 wherein said variable data relates to one of at least a sporting event score, a number of persons in a disaster or accident, an election result, a movie box office total, a number of shares traded, a medical condition, and time dependent physical variables. 16. The method of claim 1 wherein said electronic documents are based on transcriptions of speech recognized utterances. 17. The method of claim 1, wherein step (b) is performed by analyzing browsing behavior of users reviewing said first set of electronic documents. 18. The method of claim 1 wherein said steps are implemented as one or more computer software routines embodied in a tangible media and adapted to cause one or more computing systems to perform the operations recited therein. 19. The method of claim 1 wherein automated document classifier generates a temporal template for each of said one or more topics, such that a plurality of temporal states are mapped to a plurality of respective content items. 20. The method of claim 1 wherein said electronic documents are based on a transcription of an audio/video file and/or metadata associated therewith. 21. The method of claim 1 wherein an author and/or a locale associated with said documents are also identified and classified with said documents. 22. The method of claim 1 where said first temporal components for each of said one or more events and one or more topics are identified by classifying content differences between documents. 23. A method of automatically training an electronic document sorter with a computing system comprising: a) identifying a first set of set of electronic documents containing content associated with a set of topics using the computing system;wherein said content describes one or more events related to one or more topics;b) sorting said first set of electronic documents in temporal order with respect to each of said one or more events and one or more topics to form at least one sorted topic/event document set using the computing system;wherein said first set of electronic documents are sorted and annotated by a combination of both human and an automated text parsing;c) analyzing said at least one sorted topic/event document set using the computing system to identify temporal components relating to the topic/event of each electronic document which correlates to a temporal rank of such electronic document with respect to the topic/event;wherein said temporal components include temporal data including text words other than but not excluding date/time data;d) training an automated document classifier using the computing system to automatically identify second temporal components in other electronic documents associated with a particular event/topic and which are not part of said sorted topic/event document set;wherein said second temporal components also include semantic equivalents of said text words;e) collecting a new set of documents and determining a reputation/trust score based on the origin of the new set of documents;f) ranking the new set of electronic documents using said automated classifier and said reputation/trust score to determine a temporal rating of each document relative to an event or topic, which temporal rating defines a relative chronological sequence of such document for said event or topic relative to other documents in said new set; andg) selecting at least some of the documents with a highest temporal rating and outputting them in a form suitable for review by a human reviewer. 24. The method of claim 23 wherein said first temporal components include semantic equivalents. 25. The method of claim 23 wherein an author and/or a locale associated with said documents are also identified and classified with said documents. 26. The method of claim 23 where said first temporal component are identified by classifying content differences between documents. 27. The method of claim 23 wherein said steps are implemented as one or more computer software routines embodied in a tangible media and adapted to cause one or more computing systems to perform the operations recited therein.