Intelligent autofill

• US 20050257134A1
• Filed: 05/12/2004
• Published: 11/17/2005
• Est. Priority Date: 05/12/2004
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Glasgow, Dane A., Chickering, David M., Goodman, Joshua T., Kadie, Carl M., Bradford, Donald E.
• Granted Patent: US 7,660,779 B2
• Time in Patent Office: Days
• US Class Current: 715/226
• CPC Class Codes: G06F 40/174 Form filling; Merging
• CPC Class Codes: Y10S 707/99931 Database or file accessing
• CPC Class Codes: Y10S 707/99945 Object-oriented database structure processing

Abstract

The present invention provides a unique system and method that can employ machine learning techniques to automatically fill one or more fields across a diverse array of web forms. In particular, one or more instrumented tools can collect input or entries of form fields. Machine learning can be used to learn what data corresponds to which fields or types of fields. The input can be sent to a central repository where other databases can be aggregated as well. This input can be provided to a machine learning system to learn how to predict the desired outputs. Alternatively or in addition, learning can be performed in part by observing entries and then adapting the autofill component accordingly. Furthermore, a number of features of database fields as well as constraints can be employed to facilitate assignments of database entries to form values—particularly when the web form has never been seen before by the autofill system.

Claims

1. An autofill system comprising:

a trained machine learning component that uses one or more sets of inputs to facilitate autofilling of web-based forms; and

an autofill component that selectively enters appropriate data into one or more form fields based at least in part upon information generated by the machine learning component.

2. The system of claim 1, the one or more sets of inputs are stored in and retrieved from one or more databases.

3. The system of claim 2, the machine learning component learns to map from a plurality of database fields to corresponding form fields.

4. The system of claim 2, the one or more databases comprising at least one of an autofill database, contact list database, account registration database, and product registration database.

5. The system of claim 2, the machine learning component generates assignments of database entries to form fields based at least in part upon one or more constraints.

6. The system of claim 5, the one or more constraints comprising reuse constraints that limit a number of times any one database entry is used per web-based.

7. The system of claim 5, the one or more constraints comprising composition constraints that control which database entries or output types are allowed to be present concurrently on any web-based form.

8. The system of claim 5, the one or more constraints comprising constraints on fields occurring together or in a particular order.

9. The system of claim 2, the input in the database corresponds to one or more database or input fields, the database or input fields having one or more features derived therefrom.

10. The system of claim 9, at least one feature is for a particular database field being at least one of above, below, to the left of, to the right of, immediately below, immediately above, immediately to the left of, or immediately to the right of some other database field.

11. The system of claim 10, further comprising a component that determines a joint probability of a particular assignment of input fields to form fields that includes the at least one feature.

12. The system of claim 2, further comprising a component that personalizes mapping from database fields to form values by learning that for a particular user, some input field entries correspond to observed database entries and that updates the autofill component with such information.

13. The system of claim 12, the component updates the autofill component in a site-specific manner that involves leaning that some input field entries on a particular web page correspond to particular database entries.

14. The system of claim 2, the machine learning component observes that a user is entering data into a field consistent with data found in the one or more databases which triggers the autofill component to provide new autofill suggestions based in part on best ranked consistent entries.

15. The system of claim 2, the one or more databases are maintained on a machine separate from the user'"'"'s computer.

16. The system of claim 1, further comprising a data collection component that collects training data from one or more sources.

17. The system of claim 16, the data collection component aggregates the input and sends it to a central repository for further processing.

18. The system of claim 16, the data collection component comprises at least one instrumented toolbar.

19. The system of claim 1, the input comprising MAXLENGTH or SIZE entries.

20. The system of claim 1, the input comprising substrings of a field name.

21. The system of claim 1, the input comprising information from one or more adjacent form fields.

22. The system of claim 1, the input comprising at least one of nearby text or substrings of nearby text.

23. The system of claim 1, the machine learning component converts one or more inputs to one or more feature values.

24. The system of claim 23, the feature values are one of Boolean values or numeric values.

25. The system of claim 24, the Boolean values are determined automatically by performing at least one of the following:

finding useful ranges; and

looking for common and useful character sequences.

26. The system of claim 24, the Boolean values are determined manually by inspecting input values and determining one or more most useful Boolean values.

27. The system of claim 1, the autofill component provides autofill suggestions in a drop down box from a field when a user begins to enter data into the field.

28. The system of claim 1, the autofill component fills a field with suggested data as soon as a user enters the field.

29. The system of claim 1, further comprising a component tat detects whether the web-based form comprises multiple varieties of fields and separates them into at least two subsets of fields.

30. The system of claim 29, the component employs Heuristics to detect the multiple sets and to separate them into at least two subsets.

31. The system of claim 1, the machine learning component comprises a monitoring component that observes user entries of form fields to facilitate learning what value is most appropriate for the one or more form fields.

32. The system of claim 31, the machine learning component learns which of the entries a user has previously made that he is most likely to enter into a given input field.

33. The system of claim 32, the machine learning component does not learn user entries for fields that appear to comprise sensitive information.

34. The system of claim 31, the machine learning component makes use of one or more database entries and observed entries to learn at least one of the following;

a probability of a field being filled with an entry from the database times a probability that the database field is a correct entry; and

a probability of a field not being filled with an entry from the database times a probability computed by observing user entries.

35. The system of claim 31, the machine learning component generating one or more features based in part on the values possible for the fields.

36. The system of claim 35, the one or more features are order-based for each value preceding a next value.

37. The system of claim 31, the machine learning component is trained using an offline algorithm that includes collecting a plurality of observed instances of field entries and storing them.

38. The system of claim 31 the machine learning component is trained using an online algorithm using any one of Naï

ves Bayes approach, gradient descent, winnow, and perceptron algorithm.

39. The system of claim 1, the form fields comprise a file input field that displays a file based in part on recently accessed files.

40. The system of claim 39, the file entered by the autofill component is determined based in part on some combination of recently used entries in the file field and recently accessed files.

41. The system of claim 40, the combination is based at least in part on a time of last access and a time of last entry.

42. The system of claim 1, the form fields comprise radio buttons and check boxes such that the autofill component can learn whether to check or uncheck a box or select a radio button.

43. A computer readable medium having stored thereon the system of claim 1.

44. An autofill system, comprising:

a machine learning component that is trained based at least in part on observations of user entries into one or more form fields of a web-based form;

a data collection component that collects training data from one or more sources, which training data is used to train the machine learning component; and

an autofill component that automatically enters anticipated data into the one or more form fields based at least in part upon information learned from the observations made by the machine learning component.

45. An autofill system, comprising:

a trained machine learning component that generates assignments of database entries to one or more form fields of a web-based form; and

an autofill component that automatically enters the database entries into the one or more form fields based at least in part upon learned user preferences observed by the machine learning component.