REAL-TIME, STREAM DATA INFORMATION INTEGRATION AND ANALYTICS SYSTEM

• US 20170206557A1
• Filed: 06/22/2015
• Published: 07/20/2017
• Est. Priority Date: 06/23/2014
• Status: Active Grant

• Current Assignee: Board of Regents of the University of Texas System (University of Texas System)
• Original Assignee: Board of Regents of the University of Texas System (University of Texas System)
• Inventors: ABROL, Satyen, RAJASEKAR, Gunasekar, KHAN, Latifur, KHADILKAR, Vaibhav, NAGARAJAN, Siddarth, MCDANIEL, Nathan, GANESH, Gautum, THURAISINGHAM, Bhavani
• Granted Patent: US 10,902,468 B2
• Time in Patent Office: Days
• CPC Class Codes: G06Q 30/0261 based on user location
• CPC Class Codes: G06Q 30/0269 based on user profile or attribute
• CPC Class Codes: G06Q 30/0277 Online advertisement
• CPC Class Codes: G06Q 50/01 Social networking

Abstract

A real-time, stream data information integration and analytics system can include an information engine that performs real-time entity extraction to create key-value pairs of attributes for a personal profile and integrates similar personal profiles generated from same or different data sources into a single person-of-interest profile. The real-time, stream data information integration and analytics system can further include a real-time analytics module that performs a variety of analytics using the person-of-interest profiles and updates the person-of-interest profiles with scores and other results of the variety of analytics. The variety of real-time analytics can include sentiment analysis and at least some aspects of threat detection analysis. The sentiment analysis can be used to provide a recommender system and the threat detection analysis can be used to identify and predict threats and opportunities.

Claims

1. A real-time, stream data information integration and analytics system, comprising:

one or more computer-readable storage media having instructions stored thereon that, when executed by a processing system, direct the processing system to;

perform real-time entity extraction on at least a set of data within a set time window;

create key-value pairs of attributes for a personal profile using information extracted from the real-time entity extraction;

integrate personal profiles from the at least one data source that meet a specified criteria for similar personal profiles to generate person-of-interest (POI) profiles;

perform at least one real-time analytics on the POI profiles; and

update the POI profiles with results of the at least one real-time analytics.

2. The system of claim 1, wherein the instructions that direct the processing system to perform the at least one real-time analytics on the POI profiles, direct the processing system to perform a sentiment mining process.

3. The system of claim 2, wherein the one or more computer-readable storage media further comprise instructions for a recommender system stored thereon that, when executed by the processing system, direct the processing system to:

analyze messages associated with a selected POI profile for sentiment towards at least one indicated entity by performing the sentiment mining process on the messages associated with the selected POI profile; and

identify a person associated with the POI profile as a prospect for any of the at least one indicated entity having a positive sentiment result from the sentiment mining.

4. The system of claim 3, wherein the instructions for the recommender system further comprise instructions that direct the processing system to:

identify friend POI profiles for the selected POI profile;

analyze messages associated with each friend POI for sentiment towards the at least one indicated entity by performing the sentiment mining process on the messages associated with that friend POI profile;

generate a weighted vector from the determined sentiments of the friends represented by the friend POI profiles; and

in response to determining that the weighted vector meets a specified criteria, identify the person associated with the POI profile as the prospect for any of the at least one indicated entity having a positive sentiment result from the weighted vector.

5. The system of claim 3, wherein the instructions for the recommender system further comprise instructions that direct the processing system to:

communicate, to the identified person via a communication channel or social media channel, content associated with the at least one indicated entity having the positive sentiment result from the sentiment mining.

6. The system of claim 1, wherein the instructions that direct the processing system to perform the at least one real-time analytics on the POI profiles, direct the processing system to perform micro-level location mining on the set of data by at least:

performing disambiguation of locations on the set of data;

identifying general locations of the disambiguated locations; and

predicting locations for data of the set of data having no mentioned location.

7. The system of claim 1, wherein the instructions that direct the processing system to perform the at least one real-time analytics on the POI profiles, direct the processing system to perform a threat prediction by at least:

identifying threatening messages in the data set by searching for specified nouns and verbs;

identifying a corresponding POI profile for an identified threatening message and including the identified threatening message as part of an initial set of threatening messages; and

applying at least one classifier to the initial set of threatening messages to remove false positives from the initial set of threatening messages and generate an updated set of threatening messages.

8. A method comprising:

receiving a data stream from at least one data source;

performing real-time entity extraction on the data stream;

creating key-value pairs of attributes for a personal profile using information extracted from the real-time entity extraction; and

integrating personal profiles from the at least one data source that meet a specified criteria for similar personal profiles to generate person-of-interest (POI) profiles.

9. The method of claim 8, wherein creating the key-value pairs of attributes for the personal profile using the information extracted from the real-time entity extraction comprises:

organizing extracted information into key-value pairs; and

predicting values for key-value pairs having missing values.

10. The method of claim 8, wherein integrating the personal profiles from the at least one data source that meet the specified criteria for similar personal profiles to generate the POI profiles comprises:

identifying entity attributes that form a first person profile from a set of personal profiles;

identifying pairs of entity attributes, a first of the pair being from the first person profile and a second of the pair being from a data source different than that of the first person profile;

assigning scores to person profiles in the set of personal profiles based on a proximity of the pairs of entity attributes;

determining whether an assigned score of one or more of the person profiles meets the specified criteria; and

linking the first person profile to the one or more of the person profiles that are determined to have the assigned score that meets the specified criteria.

11. The method of claim 8, further comprising:

collecting a set of data from the data stream within a set time window; and

generating additional key value pairs for the personal profile based on the set of data collected within the set time window.

12. The method of claim 8, further comprising performing real-time analytics using the POI profiles;

and updating the POI profiles with results of the real-time analytics.

13. The method of claim 8, wherein performing the real-time entity extraction on the data stream comprises extracting demographic information.

14. The method of claim 13, further comprising performing sentiment mining on the data stream;

and applying a demographics-based bias.

15. The method of claim 8, further comprising performing micro-level location mining on the data stream by:

collecting a set of data from the data stream within a set time window;

performing disambiguation of locations on the set of data;

identifying general locations of the disambiguated locations; and

predicting locations for data of the set of data having no mentioned location.

16. The method of claim 8, further comprising performing a threat evaluation for at least one of the POI profiles by:

performing a demographics-based score computation to generate a demographics score for a particular POI profile;

performing a psychological evaluation score computation to generate a psychological evaluation score for the particular POI profile;

performing a content-based score computation to generate a content-based score for the particular POI profile;

performing a background check score computation to generate a background check score for the particular POI profile;

performing an online reputation-based score computation to generate a reputation score for the particular POI profile;

performing a social graph-based score computation to generate a social score for the particular POI profile; and

generating a final score for the particular POI profile by combining the demographics score, the psychological evaluation score, the content-based score, the background check score, the reputation score, and the social score for the particular POI profile.

17. The method of claim 8, further comprising performing threat prediction by:

identifying threatening messages in the data stream by searching for specified nouns and verbs;

identifying a corresponding POI profile for an identified threatening message and including the identified threatening message as part of an initial set of threatening messages;

applying at least one classifier to the initial set of threatening messages to remove false positives from the initial set of threatening messages and generate an updated set of threatening messages; and

communicating a notification of the corresponding POI profiles for the updated set of threatening messages.

18. The method of claim 17, further comprising performing micro-level location mining on the updated set of threatening messages to identify location of threats.

19. The method of claim 8, further comprising:

selecting a POI profile from the POI profiles;

analyzing messages associated with the selected POI profile to determine sentiment of a person represented by the selected POI profile towards a specified item;

in response to determining that the person has a positive sentiment towards the specified item, indicating the POI profile as a prospect for content related to the specified item.

20. The method of claim 19, further comprising:

identifying friend POI profiles for the selected POI profile;

for each friend POI profile, analyzing messages associated with that friend POI profile to determine sentiment of a friend represented by that friend POI profile towards the specified item performing sentiment mining;

generating a weighted vector from the determined sentiments of the friends represented by the friend POI profiles; and

in response to determining that the weighted vector meets a specified criteria, indicating the POI profile as the prospect for the content related to the specified item.

21. A method for a recommender system comprising:

receiving a selected person-of-interest (POI) profile, the POI profile comprising information from at least two integrated personal profiles;

receiving at least one indicated entity;

analyzing messages associated with the selected POI profile for sentiment towards the at least one indicated entity by performing a sentiment mining process on the messages associated with the selected POI profile; and

identifying a person associated with the POI profile as a prospect for any of the at least one indicated entity having a positive sentiment result from the sentiment mining.

22. The method of claim 21, further comprising:

identifying friend POI profiles for the selected POI profile;

for each friend POI profile, analyzing messages associated with that friend POI for sentiment towards the at least one indicated entity by performing the sentiment mining process on the messages associated with that friend POI profile;

generating a weighted vector from the determined sentiments of the friends represented by the friend POI profiles; and

in response to determining that the weighted vector meets a specified criteria, identifying the person associated with the POI profile as the prospect for any of the at least one indicated entity having a positive sentiment result from the weighted vector.

23. The method of claim 21, further comprising:

communicating, to a client application, at least some attributes of the POI profile for the identified person.

24. The method of claim 21, further comprising:

communicating, to the identified person via a communication channel or social media channel, content associated with the at least one indicated entity having the positive sentiment result from the sentiment mining.

25. The method of claim 21, further comprising:

identifying a physical location of the identified person and selecting at least one entity associated with the physical location as one or more of the at least one indicated entity.

26. The method of claim 25, wherein identifying the physical location of the identified person comprises performing a micro-level location mining process.

27. The method of claim 26, wherein performing the micro-level location mining process comprises:

collecting a set of data from a data stream from at least one data source within a set time window;

performing disambiguation of locations on the set of data;

identifying general locations of the disambiguated locations; and

predicting locations for data of the set of data having no mentioned location.

28. The method of claim 25, wherein identifying the physical location of the identified person comprises receiving information from a geo-enabled device of the identified person.

29. The method of claim 21, further comprising:

identifying from a database or datastream accessible over the Internet, a place of work and a residence for the person associated with the POI profile;

identifying, via a map service, at least one route between the place of work and the residence; and

identifying at least one business from which to purchase a product associated with the at least one indicated entity having the positive sentiment result from the sentiment mining.