Method and apparatus for detecting sensitive content in a document

• US 8,271,483 B2
• Filed: 09/10/2008
• Issued: 09/18/2012
• Est. Priority Date: 09/10/2008
• Status: Active Grant

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Palo Alto Research Center, Inc. (Xerox Holdings Corp.)
• Inventors: Staddon, Jessica N., Chow, Richard, de Paiva, Valeria, Golle, Philippe J. P., Fang, Ji, King, Tracy Holloway
• Primary Examiner(s): Kim, Charles
• Assistant Examiner(s): MINA, FATIMA P
• Application Number:
  US12/208,091
• Publication Number:
  US 20100076957A1
• Time in Patent Office: 1,469 Days
• Field of Search: 707/706, 707/709, 707/730, 707/736, 707/707, 707/727, 707/728
• US Class Current: 707/730
• CPC Class Codes: G06F 16/313 Selection or weighting of terms for indexing

Abstract

One embodiment of the present invention provides a system that detects sensitive content in a document. In doing so, the system receives a document, identifies a set of terms in the document that are candidate sensitive terms, and generates a combination of terms based on the identified terms that is associated with a semantic meaning. Next, the system performs searches through a corpus based on the combination of terms and determines hit counts returned for each term in the combination and for the combination. The system then determines whether the combination of terms is sensitive based on the hit count for the combination and the hit counts for the individual terms in the combination, and generates a result that indicates portions of the document which contain sensitive combinations.

Claims

1. A computer-executed method for detecting sensitive content in a document, the method comprising:

receiving a document;

identifying a set of terms in the document;

generating a combination of terms, based on the identified terms, that is associated with a semantic meaning and is potentially sensitive;

performing a first search through a public corpus for the combination of terms and determining a first search hit count returned for the combination;

performing additional searches through the public corpus for individual terms in the combination and determining search hit counts returned for each term in the combination;

computing a search hit ratio between the first search hit count and the average search hit count for the individual terms in the combination;

in response to the computed search hit ratio being smaller than a predetermined value, labeling the combination of terms as sensitive; and

generating a result that indicates portions of the document which contain sensitive combinations;

determining a relationship between the identified terms by processing the document with a syntactic parser to determine relationships between the terms;

wherein determining the relationship between the identified terms further comprises scoring a respective relationship based on a determined relevance to the document and a likelihood of the terms being sensitive.

2. The method of claim 1, wherein identifying the set of terms in the document comprises processing the document with a part-of-speech (POS) tagger to identify words that belong to a predetermined set of word classes.

3. The method of claim 1, further comprising disregarding the combination if the combination is not associated with sensitive information by determining that the respective combination can be inferred from public information.

4. The method of claim 3, wherein determining that the combination can be inferred from public information comprises:

for a respective term in the combination, gathering supplemental information related to the term based on the public corpus searches and/or synonym/hypernym databases;

determining a degree of separation between two terms in the combination by determining the relationship between a first term in the combination and the supplemental information associated with a second term in the combination; and

determining whether the degree of separation between the two terms in the combination is below a predetermined threshold.

5. A non-transitory computer-readable storage medium storing instructions that when executed by a computer cause the computer to perform a method for detecting sensitive content in a document, the method comprising:

receiving a document;

identifying a set of terms in the document;

generating a combination of terms, based on the identified terms, that is associated with a semantic meaning and is potentially sensitive;

performing a first search through a public corpus for the combination of terms and determining a first search hit count returned for the combination;

performing additional searches through the public corpus for the individual terms in the combination and determining search hit counts returned for each term in the combination;

computing a search hit ratio between the first search hit count and the average search hit count for the individual terms in the combination;

in response to the computed search hit ratio being smaller than a predetermined value, labeling the combination of terms as sensitive; and

generating a result that indicates portions of the document which contain sensitive combinations;

determining a relationship between the identified terms by processing the document with a syntactic parser to determine relationships between the terms;

wherein determining the relationship between the identified terms further comprises scoring a respective relationship based on a determined relevance to the document and a likelihood of the terms being sensitive.

6. The non-transitory computer-readable storage medium of claim 5, wherein identifying the set of terms in the document candidate comprises processing the document with a part-of-speech (POS) tagger to identify words that belong to a predetermined set of word classes.

7. The non-transitory computer-readable storage medium of claim 5, wherein the method further comprises disregarding the combination if the combination is not associated with sensitive information by determining that the respective combination can be inferred from public information.

8. The non-transitory computer-readable storage medium of claim 7, wherein determining that the combination can be inferred from public information comprises:

for a respective term in the combination, gathering supplemental information related to the term based on the public corpus searches and/or synonym/hypernym databases;

determining a degree of separation between two terms in the combination by determining the relationship between a first term in the combination and the supplemental information associated with a second term in the combination; and

determining whether the degree of separation between the two terms in the combination is below a predetermined threshold.

9. An apparatus for detecting sensitive content in a document, comprising:

a processor;

a receiving mechanism configured to receive a document;

an analysis mechanism configured to;

identify a set of terms in the document; and

generate a combination of terms, based on the identified terms, that is associated with a semantic meaning and is potentially sensitive; and

a search engine interface configured to;

perform a first search through a public corpus for the combination of terms and determine a first search hit counts returned for the combination; and

perform additional searches through the public corpus for the individual terms and determine search hit counts returned for each term in the combination;

wherein the analysis mechanism is further configured to;

compute a search hit ratio between the first search hit count and the average search hit count for the individual terms in the combination;

in response to the computed search hit ratio being smaller than a predetermined value, label the combination of terms as sensitive; and

generate a result that indicates portions of the document which contain sensitive combinations;

determine a relationship between the identified terms by processing the document with a syntactic parser to determine relationships between the terms;

wherein while determining the relationship between the identified terms, the analysis mechanism is further configured to score a respective relationship based on a determined relevance to the document and a likelihood of being sensitive.

10. The apparatus of claim 9, wherein while identifying the set of terms in the document, the analysis mechanism is further configured to process the document with a part-of-speech (POS) tagger to identify words that belong to a predetermined set of word classes.

11. The apparatus of claim 9, wherein the analysis mechanism is further configured to disregard the combination if the combination is not associated with sensitive information by determining that the respective combination can be inferred from public information.

12. The apparatus of claim 11, wherein while determining that the combination can be inferred from public information, the analysis mechanism is configured to:

for a respective term in the combination, gather supplemental information related to the term based on the public corpus searches and/or synonym/hypernym databases;

determine a degree of separation between two terms in the combination by determining the relationship between a first term in the combination and the supplemental information associated with a second term in the combination; and

determine whether the degree of separation between the two terms in the combination is below a predetermined threshold.