System and method for searching and identifying potential financial risks disclosed within a document

• US 8,180,713 B1
• Filed: 04/14/2008
• Issued: 05/15/2012
• Est. Priority Date: 04/13/2007
• Status: Active Grant

• Current Assignee: Standard & Poor's Financial Services LLC (S&P Global Inc.)
• Original Assignee: Standard & Poor's Financial Services LLC (S&P Global Inc.)
• Inventors: Feldman, Ronen, Rigby, Peter
• Primary Examiner(s): Gaffin, Jeffrey A
• Assistant Examiner(s): CHANG, LI WU
• Application Number:
  US12/102,224
• Time in Patent Office: 1,492 Days
• Field of Search: None
• US Class Current: 706/12
• CPC Class Codes: G06F 40/205 Parsing
• CPC Class Codes: G06Q 40/06 Asset management; Financial planning or analysis
• CPC Class Codes: G06Q 40/08 Insurance

Abstract

A system and method for alerting analysts to the presence of risk indicative information in a source report such as an SEC Edgar filing are described. An alert containing the risk indicative text may be e-mailed to an analyst and a link may be provided so that the analyst may easily access and view the complete report and additional identified risk indicative information. Additionally, the alerts and reports may be organized according to a hierarchical taxonomy of risk and/or corporate categories.

Claims

1. A system for alerting an analyst to the presence of risk indicative information in a source report, the system comprising:

a processor;

a content surveillance subsystem including;

at least two text analysis software modules trained with training segments extracted from historical financial reporting documents, each training segment comprising a previously observed representation of an expression of information indicative of an actual financial risk, said content surveillance subsystem being constructed and arranged to score segments of the source report based on the closeness of each segment to one or more training segments; and

a virtual committee operable to combine the scores of each of the at least two text analysis software modules; and

an alert subsystem in communication with said content surveillance subsystem and including a report generator and an alert notifier, wherein the report generator is constructed and arranged to generate a report and said the alert notifier is constructed and arranged to transmit the alert to the analyst, wherein the alert and the report comprise one or more risk indicative segments extracted from the source report by the text analysis software module where the one or more risk indicative segments in the alert are each associated with a score above a pre-determined threshold.

2. The system of claim 1, wherein said content surveillance subsystem further includes a boilerplate preprocessing module operable to identify and remove information from the source report previously determined to be irrelevant to identifying a financial risk before the text analysis software module extracts and scores risk indicative segments of the source report.

3. The system of claim 1, further comprising a taxonomy database in which each training segment is associated with at least one node in the taxonomy database and the report and the alert are organized according to the taxonomy of the one or more risk indicative segments.

4. The system of claim 1, wherein at least one of the text analysis software modules implements a text analysis algorithm comprising a classification algorithm.

5. The system of claim 4, wherein said classification algorithm comprises a maximum margin classifier.

6. The system of claim 1, wherein the alert subsystem further includes:

an alert report database;

a user database on which is stored parameters associated with users of the system; and

a user interface;

wherein the alert notifier is adapted to transmit the alert to an analyst associated with a topic of the source report as determined by parameters stored in the user database, the report generator is adapted to store the report in the alert report database, and the user interface is adapted to enable the analyst to access the report stored in the alert report database.

7. The system of claim 6, wherein the user interface comprises a web server operable to allow the user to access the report using a web browser.

8. The system of claim 6, wherein the user interface is operable to accept feedback from the analyst indicating whether a particular segment in the report is relevant or irrelevant.

9. The system of claim 1, wherein the alert is transmitted by e-mail.

10. A system for alerting an analyst to the presence of risk indicative information in a source report, the system comprising:

a processor;

a content surveillance subsystem including;

a text analysis software module trained with training segments extracted from historical financial reporting documents, each training segment comprising a previously observed representation of an expression of information indicative of an actual financial risk, said content surveillance subsystem being constructed and arranged to score segments of the source report based on the closeness of each segment to one or more training segments;

a Boolean searcher; and

a virtual committee, wherein the Boolean searcher extracts segments matching one or more pre-stored Boolean expressions and the virtual committee is operable to enhance a score of a segment extracted by the text analysis software module where the segment or a portion thereof was also extracted by the Boolean searcher; and

an alert subsystem in communication with said content surveillance subsystem and including a report generator and an alert notifier, wherein the report generator is constructed and arranged to generate a report and said the alert notifier is constructed and arranged to transmit the alert to the analyst, wherein the alert and the report comprise one or more risk indicative segments extracted from the source report by the text analysis software module where the one or more risk indicative segments in the alert are each associated with a score above a pre-determined threshold.

11. The system of claim 10, wherein said content surveillance subsystem includes at least two said text analysis software modules.

12. The system of claim 10, further comprising a taxonomy database in which each training segment and each pre-stored Boolean expression is associated with at least one node in the taxonomy database and the report and the alert are organized according to the taxonomy of the one or more risk indicative segments.

13. A computer-implemented method for automatically identifying information within a source report that is potentially indicative of a financial risk, said method comprising:

assembling a plurality of training segments from historical financial documents, each training segment comprising a previously observed representation of an expression of information indicative of an actual financial risk;

using text analysis software comprising two or more text analysis algorithms to compare content of a source report with the training segments to identify, with each text analysis algorithm, one or more portions of the content having a similarity to one or more training segments and to generate a closeness score reflective of the extent of the similarity of the identified content portion with a particular training segment;

comparing to each other the one or more portions identified with each algorithm and the closeness scores generated with each algorithm;

identifying a single identified portion having a similarity to one or more training segments and generating a single closeness score;

generating a risk alert communication for any identified portions of the content for which the generated closeness score exceeds a predetermined threshold;

transmitting the risk alert communication to at least one analyst; and

enabling the analyst to view the identified portion of the source report.

14. The method of claim 13, further comprising organizing the training segments in accordance with a hierarchical taxonomy.

15. The method of claim 13, further comprising a preprocessing step wherein content of the source report predetermined to be irrelevant to financial risk is removed prior to comparing the content of the source report with the training segments.

16. The method of claim 13, wherein the transmitting step includes transmitting by e-mail.

17. The method of claim 16, wherein the risk alert communication includes a link whereby the source report is displayed in a browser window upon the analysts selection of the link.

18. The method of claim 13, wherein the transmitting step is preceded by a step of identifying the analyst by querying a user database for the identity of an analyst associated with a topic of the source report.

19. The method of claim 13, further comprising enabling the analyst to provide feedback regarding the extent to which the content identified by the text analysis software and the closeness score reflect actual financial risks.

20. A computer-implemented method for automatically identifying information within a source report that is potentially indicative of a financial risk, said method comprising:

assembling a plurality of training segments from historical financial documents, each training segment comprising a previously observed representation of an expression of information indicative of an actual financial risk;

using text analysis software to compare content of a source report with the training segments to identify portions of the content having a similarity to one or more training segments and to generate a closeness score reflective of the extent of the similarity of the identified content portion with a particular training segment;

employing a Boolean searcher to identify segments matching one or more of a plurality of Boolean expressions stored in a database and adjusting the closeness score associated with each segment extracted by both the Boolean searcher and the text analysis software;

generating a risk alert communication for any identified portions of the content for which the generated closeness score exceeds a predetermined threshold;

transmitting the risk alert communication to at least one analyst;

enabling the analyst to view the identified portion of the source report.

21. The method of claim 20, wherein the text analysis software comprises more than one text analysis algorithm and wherein the method further comprises comparing the content of the source report with the training segment using each algorithm to identify portions of the content having a similarity to one or more training segments and to generate a closeness score reflective of the extent of the similarity of the identified content portion with a particular training segment;

comparing the portions identified and the closeness scores generated with each algorithm; and

identifying a single identified portion having a similarity to one or more training segments and generating a single closeness score.