Web browsing and machine learning systems for acquiring tax data during electronic tax return preparation

• US 10,740,854 B1
• Filed: 10/28/2015
• Issued: 08/11/2020
• Est. Priority Date: 10/28/2015
• Status: Active Grant

• Current Assignee: Intuit, Inc.
• Original Assignee: Intuit, Inc.
• Inventors: Eftekhari, Amir, Huang, Nankun
• Primary Examiner(s): Hamilton, Lalita M
• Application Number:
  US14/925,633
• Time in Patent Office: 1,749 Days
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06Q 40/123 Tax preparation or submission

Abstract

A system for acquiring tax data during preparation of an electronic tax return includes a client computer executing a browser, a browser rendering computer in communication with the client computer, and a machine learning system utilized by the browser rendering computer. The client computer is configured to present a tax preparation webpage, and receive interface data representing s first user'"'"'s interactions with a first tax data source website. The browser rendering computer is configured to execute the machine learning system, derive navigation information for the first tax data source website, generate rendering instructions for the first tax data source website, and transmit the rendering instructions to the client computer. The client computer is also configured to execute the rendering instructions to implement a second browser that presents a tax data source webpage that is controllable by a user independently of the tax preparation webpage.

Claims

1. A computerized system for acquiring tax data during preparation of an electronic tax return, the system comprising:

a client computer executing a browser and being configured to present a tax preparation webpage of a tax preparation application utilized to prepare the electronic tax return, and receive interface data from a first user for a first tax data source website of a tax data source program hosted by a tax data source computer, wherein the interface data represents the first user'"'"'s interactions with the first tax data source website;

a browser rendering computer in communication with the client computer; and

a machine learning system utilized by the browser rendering computer,

the browser rendering computer being configured to receive the interface data from the client computer, execute the machine learning system utilizing the received interface data as an input, derive navigation information for the first tax data source website based at least in part upon a result generated by the machine learning system, generate rendering instructions for the first tax data source website, and transmit the rendering instructions to the client computer, and

the client computer being further configured to execute the rendering instructions to implement a second browser that presents a tax data source webpage based at least in part upon the rendering instructions and the interface data, the client computer rendering the presented tax data source webpage such that the rendered tax data source webpage overlaps the tax preparation webpage, the tax data source webpage being controllable by a user independently of the tax preparation webpage,

the machine learning system further comprising a target classification engine configured to determine whether a second tax data source website is substantially similar to the first tax data source website and a playback engine configured to interact with the second tax data source website using the navigation information and user information provided by a second user.

2. The system of claim 1, wherein the interface data is selected from the group consisting of mouse clicks, keyboard events, form submissions, and page navigations.

3. The system of claim 1, the machine learning system comprising:

an action recorder configured to read and store the interface data; and

a navigation sequence detector configured to analyze the interface data to derive the navigation information for the first tax data source website.

4. The system of claim 3, wherein the action recorder is also configured to filter out a non-relevant event.

5. The system of claim 4, wherein the non-relevant event is selected from the group consisting of mouse clicks on non-link elements and scrolling events.

6. The system of claim 3, wherein the action recorder is also configured to read and store respective interface data from a plurality of users on the first tax data source website.

7. The system of claim 6, wherein the navigation sequence detector is also configured to analyze the respective interface data to derive the navigation information for the first tax data source website.

8. The system of claim 7, wherein analyzing the respective interface data comprises generalizing the respective interface data.

9. The system of claim 7, wherein analyzing the respective interface data comprises analyzing states of respective tax data source webpages of the first tax data source website resulting from applying the respective interface data to the first tax data source website.

10. The system of claim 3, wherein the navigation sequence detector comprises a branch detector configured to analyze first and second tax data source webpages of the first tax data source website, wherein the first and second tax data source webpages have different configurations but respective interactions on the first and second tax data source webpages result in respective first and second result webpages having the same state.

11. The system of claim 1, wherein the target classification engine is also configured to convert each of the first and second tax data source websites to respective first and second tree structures, and to perform an edit-distance computation on the first and second tree structures to determine whether the first and second tax data source websites are substantially similar to each other.

12. The system of claim 11, wherein the target classification engine is also configured to categorize the first and second tax data source websites as substantially similar when the edit-distance computation outputs a difference parameter greater than a predetermined difference parameter.

13. The system of claim 1, wherein the playback engine is also configured to enter the user information into a second field in the second tax data source website using the navigation information, wherein the second field in the second tax data source website corresponds to a first field in the first tax data source website.

14. The system of claim 1, the machine learning system further comprising a parameter detector configured to parameterize a field on the second tax data source website before the playback engine interacts with the second tax data source website.

15. The system of claim 14, wherein parameterizing the field on the second tax data source website comprises determining that the field includes a parameter.

16. The system of claim 15, wherein the parameter is selected from the group consisting of an HTTP POST request and an onblur event.

17. The system of claim 15, wherein parameterizing the field on the second tax data source website further comprises capturing the parameter.

18. The system of claim 17, wherein parameterizing the field on the second tax data source website further comprises removing the parameter from the second tax data source website before the playback engine interacts with the second tax data source website.

19. A computer-implemented method of acquiring tax data using a computerized system comprising a client computer executing a browser, a browser rendering computer in communication with the client computer, and a machine learning system utilized by the browser rendering computer, the method comprising:

the client computer presenting a tax preparation webpage of a tax preparation application utilized to prepare the electronic tax return;

the client computer receiving interface data from a first user for a first tax data source website of a tax data source program hosted by a tax data source computer, wherein the interface data represents the first user'"'"'s interactions with the first tax data source website;

the browser rendering computer receiving the interface data from the client computer;

the browser rendering computer executing the machine learning system utilizing the received interface data as an input;

the browser rendering computer deriving navigation information for the first tax data source website based at least in part upon a result generated by the machine learning system;

the browser rendering computer generating rendering instructions for the first tax data source website, and transmitting the rendering instructions to the client computer; and

the client computer executing the rendering instructions to implement a second browser that presents a tax data source webpage based at least in part upon the rendering instructions and the interface data, the client computer rendering the presented tax data source webpage such that the rendered tax data source webpage overlaps the tax preparation webpage, the tax data source webpage being controllable by a user independently of the tax preparation webpage,

wherein the machine learning system further comprises a target classification engine and a playback engine, and said method further comprises the target classification engine determining whether a second tax data source web site is substantially similar to the first tax data source website and the playback engine interacting with the second tax data source website using the navigation information and user information provided by a second user.

20. The method of claim 19, wherein the interface data is selected from the group consisting of mouse clicks, keyboard events, form submissions, and page navigations.

21. The method of claim 19, wherein the machine learning system comprises an action recorder and a navigation sequence detector, the method further comprising:

the action recorder reading and storing the interface data; and

the navigation sequence detector analyzing the interface data and deriving the navigation information for the first tax data source website.

22. The method of claim 21, further comprising the action recorder filtering out a non-relevant event.

23. The method of claim 22, wherein the non-relevant event is selected from the group consisting of mouse clicks on non-link elements and scrolling events.

24. The method of claim 21, further comprising the action recorder reading and storing respective interface data from a plurality of users on the first tax data source website.

25. The method of claim 24, further comprising the navigation sequence detector analyzing the respective interface data and deriving the navigation information for the first tax data source website.

26. The method of claim 25, wherein analyzing the respective interface data comprises the navigation sequence detector generalizing the respective interface data.

27. The method of claim 25, wherein analyzing the respective interface data comprises the navigation sequence detector analyzing states of respective tax data source webpages of the first tax data source website resulting from applying the respective interface data to the first tax data source website.

28. The method of claim 21, wherein the navigation sequence detector comprises a branch detector, the method further comprising the branch detector analyzing first and second tax data source webpages of the first tax data source website, wherein the first and second tax data source webpages have different configurations but respective interactions on the first and second tax data source webpages result in respective first and second result webpages having the same state.

29. The method of claim 19, further comprising the target classification engine converting each of the first and second tax data source websites to respective first and second tree structures, and performing an edit-distance computation on the first and second tree structures to determine whether the first and second tax data source websites are substantially similar to each other.

30. The method of claim 29, further comprising the target classification engine categorizing the first and second tax data source websites as substantially similar when the edit-distance computation outputs a difference parameter greater than a predetermined difference parameter.

31. The method of claim 19, further comprising the playback engine entering the user information into a second field in the second tax data source website using the navigation information, wherein the second field in the second tax data source website corresponds to a first field in the first tax data source website.

32. The method of claim 19, wherein the machine learning system further comprises a parameter detector, the method further comprising the parameter detector parameterizing a field on the second tax data source website before the playback engine interacts with the second tax data source website.

33. The method of claim 32, wherein parameterizing the field on the second tax data source website comprises determining that the field includes a parameter.

34. The method of claim 33, wherein the parameter is selected from the group consisting of an HTTP POST request and an onblur event.

35. The method of claim 33, wherein parameterizing the field on the second tax data source website further comprises capturing the parameter.

36. The method of claim 35, wherein parameterizing the field on the second tax data source website further comprises removing the parameter from the second tax data source website before the playback engine interacts with the second tax data source website.