System and method for research report guided proactive news analytics for streaming news and social media

• US RE46,983 E1
• Filed: 02/06/2017
• Issued: 08/07/2018
• Est. Priority Date: 06/25/2013
• Status: Active Grant

• Current Assignee: JP Morgan Chase Bank, N.A. (JP Morgan Chase & Co)
• Original Assignee: JP Morgan Chase Bank, N.A. (JP Morgan Chase & Co)
• Inventors: Kursun, Eren, Fernandez, Gene, Berson, Alex, Goodman, Brian
• Primary Examiner(s): Wood, William H
• Application Number:
  US15/425,726
• Time in Patent Office: 547 Days
• CPC Class Codes: G06F 16/24568 Data stream processing; Continuous queries
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 16/367 Ontology
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06F 17/2235 Hyperlinking information retrieval based on hyperlinks G06F17/30014
• CPC Class Codes: G06F 17/2765 Recognition
• CPC Class Codes: G06F 17/2785 Semantic analysis
• CPC Class Codes: G06F 17/28 Processing or translating of natural language G06F17/27 takes precedence
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

Systems and methods may provide proactive news analytics based on integrated prediction statements. Users may extract and collect conditional statements from research reports. A compiled list based on the processing/linking of statements for signal generation may then be created. Similarly, a list of counter statements that are assigned a conflict rating that specifies how much agreement/disagreement on a specific topic exists for the counter statement and the statement itself may be created. The custom library may have semantic capabilities to justify conditional statements in order to capture meaning and identify supporting news related to the statement. When a relevant event is detected that relates to a conditional statement, expected conclusions are linked, and customized indexes are calculated to allow for analysis of the relevant event.

Claims

1. A computer implemented method, comprising:

extracting, by a computer processor, one or more statements from one or more unstructured sources wherein the extracted statements comprise predictive and conditional statements pertaining to an event in the future;

normalizing, by the computer processor, the extracted statements;

linking the extracted statements from the one or more unstructured sources based on one or more factors;

testing the extracted statements for consistency, wherein inconsistent statements are removed;

creating, by the computer processor, a list of custom keywords based on the extracted statements;

ranking the extracted statements based on a strength of agreement; and

creating an alternative conditional statement flow chart upon the extracted statements having a high degree of disagreement.

2. The method of claim 1, further comprising:

associating a sentiment position signal with each of the extracted statements;

creating a custom keyword library and natural language processing structures based on each sentiment position signal;

linking the custom keyword library and the natural language processing structures based on a semantic analysis;

prefetching data from a stream of news, social media, and unstructured sources wherein the data is prefetched based on potential matching of the data to items in the custom keyword library and the natural language processing structures;

analyzing the prefetched data based on the custom keyword library and the natural language processing structures for matching items;

processing the matching items based upon sentiment and confidence scoring; and

creating a customized output.

3. The method of claim 2, wherein the potential matching of the data to items in the custom keyword library and the natural language processing structures is measured at least in terms comprising one or more of:

entity matching, keyword matching, and timeline matching.

4. The method of claim 2, further comprising:

verifying the matching items using a set of cross-channel checks that comprise verifying each matching item against one or more other independent sources.

5. The method of claim 2, wherein the stream is a weighted stream with weights applied to each source in the stream.

6. The method of claim 2, further comprising:

implementing machine based learning for future prefetching based on an accuracy of the matching items and the event.

7. The method of claim 2, further comprising:

applying a feedback mechanism that extracts information from the one or more unstructured sources pertaining to the occurrence of the event.

8. The method of claim 1, wherein the unstructured sources comprise research reports.

9. The method of claim 1, wherein the one or more factors comprise one or more of:

the conditions or predictions that the extracted statements refer to, an expected timeline, an expected cause, an expected dependency, a keyword or keywords, and a list of tracked entities.

10. The method of claim 1, further comprising:

creating a linked and timeline based flow diagram based on the extracted statements and the event.

11. A system, comprising:

a processor; and

a memory comprising computer-readable instructions which when executed by the processor cause the processor to perform the steps comprising;

extracting one or more statements from one or more unstructured sources wherein the extracted statements comprise predictive and conditional statements pertaining to an event in the future;

normalizing the extracted statements;

linking the extracted statements from the one or more unstructured sources based on one or more factors;

testing the extracted statements for consistency, wherein inconsistent statements are removed;

creating a list of custom keywords based on the extracted statements;

ranking the extracted statements based on a strength of agreement; and

creating an alternative conditional statement flow chart if the extracted statements have a high degree of disagreement.

12. The system of claim 11, further comprising:

associating a sentiment position signal with each of the extracted statements;

creating a custom keyword library and natural language processing structures based on each sentiment position signal;

linking the custom keyword library and the natural language processing structures based on a semantic analysis;

prefetching data from a stream of news, social media, and unstructured sources wherein the data is prefetched based on potential matching of the data to items in the custom keyword library and the natural language processing structures;

analyzing the prefetched data based on the custom keyword library and the natural language processing structures for matching items;

processing the matching items based upon sentiment and confidence scoring; and

creating a customized output.

13. The system of claim 12, further comprising:

verifying the matching items using a set of cross-channel checks that comprise verifying each matching item against one or more other independent sources.

14. The system of claim 12, wherein the stream is a weighted stream with weights applied to each source in the stream.

15. The system of claim 12, further comprising:

implementing machine based learning for future prefetching based on an accuracy of the matching items and the event.

16. The system of claim 12, further comprising:

applying a feedback mechanism that extracts information from the one or more unstructured sources pertaining to the occurrence of the event.

17. The system of claim 11, wherein the unstructured sources comprise research reports.

18. The system of claim 11, wherein the one or more factors comprise one or more of:

the conditions or predictions that the extracted statements refer to, an expected timeline, an expected cause, an expected dependency, a keyword or keywords, and a list of tracked entities.

19. The system of claim 11, further comprising:

creating a linked and timeline based flow diagram for the extracted statements and the event.

20. A computer implemented method, comprising:

acquiring, by a computer processor, a plurality of statements from multiple input channels, wherein each of the plurality of statements relates to an event that is predicted to occur in the future and each of the multiple input channels has an inherent confidence score based on historical performance of each of the multiple input channels relating to statements and predicted events;

processing, by the computer processor, each of the plurality of statements, wherein the processing comprises one or more of;

linking data structures with time stamps for action, assigning importance, and delineating consequences;

normalizing each of the plurality of statements to account for differences in structure and format of the multiple input channels;

linking and compiling, by the computer processor, each of the plurality of statements through one or more of hierarchy, timeline, entity connectivity, and causality implications;

verifying each of the plurality of statements for consistency in a multi-dimensional space;

removing inconsistent statements that fall below an established threshold for consistency;

calculating, by the computer processor, a unique confidence score based on a matching between two or more of the plurality of statements;

forming a new group of statements based on remaining statements, wherein the new group of statements is weighted based on the inherent confidence score;

creating, by the computer processor, a custom keyword library based on the new group of statements;

creating, by the computer processor, natural language processing structures for statement justification;

associating sentiment signals with the new group of statements;

linking, by the computer processor, the custom keyword library and the natural language processing structures;

prefetching signals, by the computer processor, from a stream comprising one or more sources comprising news, social media, and research reports;

checking, by the computer processor, each prefetched signal for a potential match to the custom keyword library and the natural language processing structures, wherein upon a determination of a match the potential match becomes a matched signal; and

performing, by the computer processor, a confidence check on each matched signal based on the weighting based on the inherent confidence score.

21. The computer implemented method of claim 20, wherein the multiple input channels comprise:

user inputs, text tear outs from documents, automatically extracted statements from documents, news sources, and inputs from individual analysts.

22. The computer implemented method of claim 21, wherein the inputs from individual analysts comprise machine readable data comprising one or more of flow charts, quantitative models, predictive statements, or a combination thereof.

23. The computer implemented method of claim 20, wherein the multi-dimensional space comprises time, causality, and hierarchy.

24. The computer implemented method of claim 20, further comprising:

normalizing each word in the new group of statements; and

compiling a statement list with analog words and statements based on the new group of statements; and

adding the statement list to the custom keyword library.

25. The computer implemented method of claim 20, wherein the natural language processing structures identify analogous expressions to contents of the custom keyword library.

26. The computer implemented method of claim 20, further comprising:

retroactively scanning the stream, after the event has occurred, for signals related to occurrence of the event;

processing, by a computer processor, the signals related to the occurrence;

performing, by the computer processor, a statistical analysis of the signals;

identifying a custom library of keywords from the statistical analysis;

storing the custom library of keywords from the statistical analysis;

implementing machine learning based on the custom library of keywords and the occurrence to improve future prefetching of signals.

27. A computer implemented method, comprising:

acquiring, by a computer processor, a plurality of statements related to a specific entity from multiple input channels, wherein each of the plurality of statements relates to an event that is predicted to occur in the future and each of the multiple input channels has an inherent confidence score based on historical performance associated therewith;

processing, by the computer processor, each of the plurality of statements, wherein the processing comprises one or more of;

linking data structures with time stamps for action, assigning importance, and delineating consequences;

normalizing each of the plurality of statements to account for differences in structure and format of the multiple input channels;

linking and compiling, by the computer processor, each of the plurality of statements through one or more of hierarchy, timeline, entity connectivity, and causality implications;

verifying each of the plurality of statements for consistency in a multi-dimensional space;

removing inconsistent statements falling below an established threshold of consistency;

calculating, by the computer processor, a unique confidence score based on a matching between two or more of the plurality of statements;

forming a new group of statements based on remaining statements;

creating, by the computer processor, a custom keyword library based on the new group of statements;

creating, by the computer processor, natural language processing structures for statement justification;

associating sentiment signals with the new group of statements;

linking, by the computer processor, the custom keyword library and the natural language processing structures;

prefetching signals, by the computer processor, from a stream comprising one or more sources comprising news, social media, and research reports;

weighting each prefetched signal based on the inherent confidence score;

checking, by the computer processor, each prefetched signal for a potential match to the custom keyword library and the natural language processing structures, wherein upon a determination of a match the potential match becomes a matched signal;

performing, by the computer processor, a confidence check comprising assignment of a confidence score on each matched signal based on the weighting; and

combining each matched signal to create a position signal for the specific entity using the weighting.

28. The computer implemented method of claim 27, wherein the multiple input channels comprise:

user inputs, text tear outs from documents, automatically extracted statements from documents, news sources, and inputs from individual analysts.

29. The computer implemented method of claim 27, wherein the multi-dimensional space comprises time, causality, and hierarchy.

30. A computer implemented method, comprising:

acquiring, by a computer processor, a plurality of statements from multiple input channels, wherein each of the plurality of statements relates to an event that is predicted to occur in the future and each of the multiple input channels has an inherent confidence score based on historical performance associated therewith;

processing, by the computer processor, each of the plurality of statements, wherein the processing comprises one or more of;

linking data structures with time stamps for action, assigning importance, and delineating consequences;

normalizing each of the plurality of statements to account for differences in structure and format of the multiple input channels;

linking and compiling, by the computer processor, each of the plurality of statements through one or more of hierarchy, timeline, entity connectivity, and causality implications;

verifying each of the plurality of statements for consistency in a multi-dimensional space;

calculating, by the computer processor, a unique confidence score based on a matching between two or more of the plurality of statements;

removing inconsistent statements based on an established threshold, wherein statements below the threshold are removed;

forming a new group of statements based on remaining statements following the removing;

creating, by the computer processor, a custom keyword library based on the new group of statements;

creating, by the computer processor, natural language processing structures for statement justification;

associating sentiment signals with the new group of statements;

linking, by the computer processor, the custom keyword library and the natural language processing structures;

prefetching signals, by the computer processor, from a stream comprising one or more sources comprising news, social media, and research reports;

checking, by the computer processor, each prefetched signal for a potential match to the custom keyword library and the natural language processing structures, wherein upon a determination of a match the potential match becomes a matched signal;

applying, by the computer processor, machine learning based on each matched signal, wherein the machine learning comprises identifying indicators in the matched signal to identify additional signals to pre-fetch; and

prefetching, from the stream, by the computer processor, one or more additional signals based on the machine learning.

31. The computer implemented method of claim 30, wherein the multiple input channels comprise:

user inputs, text tear outs from documents, automatically extracted statements from documents, news sources, and inputs from individual analysts.

32. The computer implemented method of claim 30, wherein the multi-dimensional space comprises time, causality, and hierarchy.