Back to results
US 9,177,254 B2Granted
Event detection through text analysis using trained event template models
AssigneeQbase, LLC
InventorsSanjay Boddhu, Robert Flagg, Rakesh Dave, Scott Lightner
Life of this patent
Estimating the term…
Abstract
A system and method for detecting events based on input data from a plurality of sources. The system may receive input from a plurality of sources containing information about possible events. A method for event detection involves pre-processing and normalizing a data input from a plurality of sources, extracting and disambiguating events and entities, associate event and entities, correlate events and entities associated from a data input to results from a different data sources to determine if an event has occurred, and store the detected events in a data storage.
Claims14
A computer-implemented method comprising:
- identifying, by the computer, in a data stream received from a data source among a plurality of data sources one or more features matching a model feature in an event model stored in non-transitory machine-readable storage of an event concept store, wherein at least one feature in the one or more features is an event candidate corresponding to the event model, and wherein at least one feature in the one or more features is an entity;
- assigning, by the computer, a weight to each respective entity according to the event model corresponding to the event candidate;
- calculating, by the computer, an event probability score based on one or more weights respectively assigned to the one or more entities;
- associating, by the computer, the event candidate with each respective entity into a first record of a spatial-temporal grouping, in response to the computing device determining that the event probability score satisfies an event likelihood threshold score of the event model;
- storing, by the computer, the first record containing the event candidate, each respective entity associated with the event candidate, and the event probability score into a non-transitory machine-readable spatial-temporal grouping memory, wherein the spatial-temporal memory stores a plurality of records of spatial-temporal groupings, wherein each respective record is associated with each respective data source, whereby each respective record comprises the event candidate, a set of one or more entities identified in the associated data source, and the event probability score for each respective record;
- calculating, by the computer, a validation score for the first record based on a comparison of each of the event probability scores in each respective record containing the event candidate;
- and
- storing, by the computer, the event candidate of the first record as a new verified event into non-transitory machine-readable storage memory of a verified event store, in response to determining the validation score for the first record satisfies a validation threshold.
5 more elements
A system for detecting and extracting events comprising:
- an event concept store comprising non-transitory machine-readable memory storing one or more event models, wherein an event model corresponds to an event candidate and comprises a threshold event score and a set of one or more features having a corresponding weight;
- an entity and topic extraction processor configured to extract a set of entities and a set of topics from a data stream and then disambiguate each topic and each entity;
- an event extraction processor configured to identify each of the features of each event model that occur in the data stream, calculate an event score for one or more event candidates having an identified feature using the corresponding event model, and then extract the event candidate when an event score satisfies the threshold event score of the event model;
- a spatial-temporal event processor configured to associate each topic and entity extracted from each of the data streams with each of the event candidates extracted from each of the data streams to form a spatial-temporal event grouping comprising one or more records, wherein a record is the event candidate and the associated topic or entity of a data stream;
- a verified event store storing one or more verified events;
- and
- an event validation processor for verifying that each of the event candidates correspond to a real-world event and is configured to calculate a verification score for each event candidate in the spatial-temporal grouping based on a comparison of each of the records in the spatial-temporal grouping, and then store the event candidate as a verified event in the verified event store when the event candidate satisfies a verification threshold.
4 more elements
Dependencies are inferred from the claim text.