Text searching and categorization tools

• US 7,996,356 B2
• Filed: 03/24/2005
• Issued: 08/09/2011
• Est. Priority Date: 03/24/2005
• Status: Expired due to Fees

• Current Assignee: Realm Licensing LLC (IP Edge LLC)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Martin, Nathaniel G.
• Primary Examiner(s): Pulliam; Christyann R
• Application Number:
  US11/088,513
• Publication Number:
  US 20060218105A1
• Time in Patent Office: 2,329 Days
• Field of Search: None
• US Class Current: 707/603
• CPC Class Codes: G06N 5/025 Extracting rules from data

Abstract

A disclosed process accesses text data that is to be mined. The text data includes text snippets. Rules are encoded in a rule base. A search request is submitted to a search request handler. A search request handler applies the rules from the rule base to the text and associates different labels to respective text snippets in the text data in accordance with the rule base.

Claims

1. Machine-readable media encoded with data, the data being interoperable with a machine to cause:

accessing text data to be mined, the text data including text snippets;

encoding rules in a rule base, the encoding of a given one of the rules including a human user entering, via a computer screen displayed subject matter expert interface, freely typed text representing a given label and the encoding of the given one further including the human user entering, via a computer screen interface, freely typed text representing given synonyms, the given label and the given synonyms defining the given one of the rules at least in part;

submitting a search request to a search request handler;

the search request handler applying the rules from the rule base, including the given one, to the text data and automatically associating different labels to respective text snippets in the text data in accordance with the rule base;

displaying the text snippets and associated labels resulting from the application of the rules from the rule base on the subject matter expert interface; and

after the search request handler has at least once applied the rules, presenting to the human user, on the computer screen displayed subject matter expert interface, a revise option that can be selected by the human user through the subject matter expert interface to indicate a need to further encode the rules in the rule base, and the human user choosing the revise option and entering, via the computer screen displayed subject matter expert interface, freely typed text to thereby revise both the given label and the given synonyms;

wherein the rule input by the human user includes information encoding a rule from among the rules in the rule base to include a label and synonyms including a corresponding set of match terms, where a mined text snippet containing a match term in the corresponding set of match terms is associated with the label.

2. The machine-readable media according to claim 1, the data being interoperable with a machine to further cause:

after presenting the human user with text snippets and their associated different labels, in response to the human user'"'"'s selection of the revise option, further encoding further encoded rules in the rule base, the further encoded rules including one or both of modified versions of the rules and added rules.

3. The machine-readable media according to claim 2, the data being interoperable with a machine to further cause:

the encoding and the further encoding being in response to rule information input by the human user via the computer screen displayed subject matter expert interface.

4. The machine-readable media according to claim 3, the data being interoperable with a machine to further cause:

the encoding being in response to rule information input by a human user via the computer screen displayed subject matter expert interface, the rule information input by the human user via the subject matter expert interface including both the human user entering of the freely typed text representing the given label and the human user entering of the freely typed text representing the given synonyms; and

the rule information input by the user including a textual language using a prescribed syntax, the textual language including a textual label separated from match terms and from negation terms with a prescribed symbol.

5. The machine-readable media according to claim 3, the data being interoperable with the machine to further cause:

a rule from among the rules in the rule base being encoded in response to input by the human user by the computer screen displayed subject matter expert interface to also include a corresponding set of negation terms, and where a mined text snippet containing at least one negation term in a given set of negation terms is not associated with the label corresponding to the given set of negation terms.

6. The machine-readable media according to claim 5, the data being interoperable with the machine to further cause:

a rule from among the rules in the rule base further including, for respective terms of the rule, corresponding sets of synonyms, and where a mined text snippet contains a term when it contains a string of characters matching the term'"'"'s string of characters or a corresponding synonym'"'"'s string of characters.

18. The machine-readable media according to claim 1, the data being interoperable with a machine to further cause:

presenting to the human user, via the computer screen displayed subject matter expert user interface, a cluster adequacy indication graphical tool;

the search request handler awaiting a cluster adequacy indication input by the human user, via the cluster adequacy indication graphical tool on the computer screen displayed subject matter expert interface, regarding whether given clusters, resulting from the search request handler having applied the rules from the rule base to a given set of mined text data, are adequate, and, when the awaited cluster adequacy indication graphical tool by the human user is that the given clusters are adequate, the search request handler subsequently performing a post-search analysis on the given set of mined text data, and, when the awaited cluster adequacy indication is that the given clusters are inadequate, subsequently receiving, from the human user via the computer screen displayed subject matter expert user interface, additional rule encoding information for the given set of mined text data and requesting the search request handler to search the given set of mined text data using the additional rule encoding information.

19. The machine-readable media according to claim 18, the data being interoperable with a machine to further cause:

the cluster adequacy indication to be received from a subject matter expert input displayed on the computer screen and interacted with by the human user.

20. The machine-readable media according to claim 18, the data being interoperable with a machine to further cause:

the cluster adequacy indication to be received from the human user via a computer screen displayed search interface forming part of the subject matter expert interface.

21. The machine-readable media according to claim 18, the data being interoperable with a machine to further cause:

presenting to the human user via the computer screen displayed subject matter expert user interface the cluster adequacy graphical tool, the cluster adequacy graphical tool being configured to receive, via the computer screen displayed interface, one of adequate and not adequate inputs from the human user, thereby causing the cluster adequacy indication awaited by the search request handler.

7. A method comprising:

accessing text data to be mined, the text data including text snippets;

encoding rules in a rule base, the encoding of a given one of the rules including a user entering, via a computer screen displayed subject matter expert interface, freely typed text representing a given label and the encoding of the given one further including the user entering, via the computer screen displayed subject matter expert interface, freely typed text representing given synonyms, the given label and the given synonyms defining the given one of the rules at least in part;

submitting a search request to a search request handler;

the search request handler applying the rules from the rule base to the text data and associating different labels to respective text snippets in the text data in accordance with the rule base;

displaying the text snippets and associated labels resulting from the application of the rules from the rule base on the subject matter expert interface; and

after the search request handler has at least once applied the rules, presenting to the human user, on the computer screen displayed subject matter expert interface, a revise option that can be selected by the human user through the subject matter expert interface to indicate a need to further encode the rules in the rule base, and the human user choosing the revise option and entering, via the computer screen displayed subject matter expert interface, freely typed text to thereby revise both the given label and the given synonyms;

wherein the rule input by the human user includes information encoding a rule from among the rules in the rule base to include a label and synonyms including a corresponding set of match terms, where a mined text snippet containing a match term in the corresponding set of match terms is associated with the label.

8. The method according to claim 7, further comprising:

after presenting the human user with text snippets and their associated different labels, in response to the human user'"'"'s selection of the revise option, further encoding rules in the rule base, the further encoded rules including one or both of modified versions of the rules and added rules.

9. The method according to claim 8, further comprising:

the encoding and the further encoding being in response to rule information input by the human user via the computer screen displayed subject matter expert interface.

10. The method according to claim 9, further comprising:

the encoding being in response to rule information input by the human user via a computer screen displayed rule selection interface forming part of the subject matter expert interface, the rule selection interface including graphical tools configured to allow the human user to add rules, and to select one or more rules from a rule library; and

the rule information input by the user including a textual language using a prescribed syntax, the textual language including a textual label separated from match terms and from negation terms with a prescribed symbol.

11. The method according to claim 9, further comprising:

a rule from among the rules in the rule base being encoded in response to input by the human user by the computer screen displayed subject matter expert interface to also include a corresponding set of negation terms, and where a mined text snippet containing at least one negation term in a given set of negation terms is not associated with the label corresponding to the given set of negation terms.

12. The method according to claim 11, further comprising:

a rule from among the rules in the rule base further including for respective terms of the rule corresponding sets of synonyms, and where a mined text snippet contains a term when it contains a string of characters matching the term'"'"'s string of characters or a corresponding synonym'"'"'s string of characters.

13. Apparatus comprising:

a computer screen displaying a computer screen interface;

a search request handler to access text data to be mined, the text data including text snippets;

a rule base including rules;

a rule entering interface configured to receive a given one of the rules input by a human user, via a rules selection interface displayed on the computer screen, freely typing text representing a given label and further by the human user, via the computer screen interface, freely typing text representing given synonyms, the given label and the given synonyms defining the given one of the rules at least in part;

the rule selection interface to encode, via the human user interacting with the computer screen, the rules in the rule base;

a search interface to submit a search request to the search request handler; and

an execution mechanism to apply the rules from the rule base, including the given one, to the text data and to associate different labels to respective text snippets in the text data in accordance with the rule base;

a subject matter expert interface to display the text snippets and associated labels resulting from the application of the rules from the rule base; and

a new rule input mechanism preconfigured to present via the computer screen displayed rule selection interface, after the execution mechanism has at least once applied the rules, a revise option that can be selected by the human user through the subject matter expert interface to indicate a need to further encode the rules in the rule base, and to receive, upon the human user choosing the revise option, freely typed text to thereby revise both the given label and the given synonyms;

wherein a rule in the rule base includes a label and synonyms including a corresponding set of match terms, where, when the rule is executed by the execution mechanism, a mined text snippet containing at least one match term in the given set of match terms is associated with the label corresponding to the given set of match terms.

14. The apparatus according to claim 13, further comprising:

a results display to present to the human user, on the computer screen display, text snippets and their associated different labels.

15. The apparatus according to claim 14, further comprising:

a rule modification interface to modify rules in the rule base and to add rules to the rule base.

16. The apparatus according to claim 15, wherein:

the rules stored in the rule base include textual language using a prescribed syntax, the textual language including a textual label separated from match terms and from negation terms with a prescribed symbol.

17. The apparatus according to claim 15, wherein:

the rule in the rule base also includes a corresponding set of negation terms, and where, when the rule is executed by the execution mechanism, a mined text snippet containing at least one negation term in a given set of negation terms is not associated with the label corresponding to the given set of negation terms.