Back to results
US 8,078,625 B1Granted
URL-based content categorization
AssigneeAOL Inc.
InventorsJianping Zhang, Jinshui Qin, Qiuming Yan
Life of this patent
Estimating the term…
Abstract
Content may be categorized by accessing a URL associated with the content, determining a set of n-grams contained in the URL, and determining a category of the content based on the set of n-grams.
Claims31
A computer implemented method of categorizing content, the method comprising:
- accessing a uniform resource locator associated with content requested by a user;
- processing the uniform resource locator to generate a set of multiple n-grams contained in the uniform resource locator, wherein the processing includes using a plurality of different n values to generate the set of multiple n-grams, the set including n-grams of different character lengths;
- providing the set of multiple n-grams to a classifier for determination of a classification score based on the set of multiple n-grams, wherein each of the n-grams is weighted according to the formula:
- R
-
- (
- t
- )
- =
- P
-
- (
- t
- ❘
- C
- )
- P
-
- (
- t
- ❘
- C
- _
- )
- +
- r
- *
- r
- wherein t is an n-gram, C is a target category, C is a non-target category, and r is a parameter;
- comparing the classification score to a threshold;
- classifying the content into the target category or the non-target category based on the comparison of the classification score to the threshold;
- allowing the user to access the content when the content is classified into the non-target category;
- and
- blocking access to the content by the user when the content is classified into the target category.
31 more elements
A computer implemented method of categorizing content, the method comprising:
- accessing, from computer storage, a uniform resource locator associated with content;
- determining a set of n-grams contained in the uniform resource locator, wherein a plurality of different n values are used to determine the set of n-grams, the set of n-grams including n-grams of different character lengths;
- determining a category of the content based on the set of n-grams, wherein each of the n-grams is weighted according to the formula:
- R
-
- (
- t
- )
- =
- P
-
- (
- t
- ❘
- C
- )
- P
-
- (
- t
- ❘
- C
- _
- )
- +
- r
- *
- r
- wherein t is an n-gram, C is a target category, C is a non-target category, and r is a parameter;
- and
- storing, in the computer storage, an association between the determined category and the content.
28 more elements
A system for categorizing content, the system comprising:
- at least one processing device;
- a storage storing instructions for causing the at least one processing device to implement:
- an n-gram generator configured to access a uniform resource locator (URL) associated with content and determine a set of n-grams contained in the uniform resource locator;
- a classifier configured to determine a URL score based on the set of n-grams, wherein each of the n-grams is weighted according to the formula:
- R
-
- (
- t
- )
- =
- P
-
- (
- t
- ❘
- C
- )
- P
-
- (
- t
- ❘
- C
- _
- )
- +
- r
- *
- r
- wherein t is an n-gram, C is a target category, C is a non-target category, and r is a parameter that favors the n-gram based on a frequency and distinctness of the n-gram;
- and
- a categorizer configured to categorize the content based on the URL score.
29 more elements
A system for categorizing content, the system comprising:
- at least one processing device;
- a storage storing instructions for causing the at least one processing device to implement:
- a URL-based classifier configured to produce a first score based on a set of n-grams contained in a uniform resource locator associated with content, wherein the set of n-grams is based on multiple different n-values and includes n-grams of different character lengths, wherein each of the n-grams is weighted according to the formula:
- R
-
- (
- t
- )
- =
- P
-
- (
- t
- ❘
- C
- )
- P
-
- (
- t
- ❘
- C
- _
- )
- +
- r
- *
- r
- wherein t is an n-gram, C is a target category, C is a non-target category, and r is a parameter that favors the n-gram based on a frequency and distinctness of the n-gram;
- a content-based classifier configured to produce a second score based on the content, wherein the second score is separate from the first score;
- and
- an integration engine configured to determine a category of the content based on both the first score and the second score.
29 more elements
The system of 21 wherein, to determine a category of the content based on both the first score and the second score, the integration engine is configured to generate a sum of the first score and the second score and compare the sum to a threshold.
A computer implemented method of categorizing content, the method comprising:
- accessing a uniform resource locator (URL) associated with the content;
- processing the uniform resource locator to generate a set of multiple n-grams contained in the uniform resource locator, wherein a plurality of different n values are used to process the uniform resource locator, and the set of multiple n-grams includes n-grams of different character lengths;
- providing the set of multiple n-grams to a URL classifier, wherein the URL classifier is configured to determine a first classification score based on the set of multiple n-grams, wherein each of the n-grams is weighted according to the formula:
- R
-
- (
- t
- )
- =
- P
-
- (
- t
- ❘
- C
- )
- P
-
- (
- t
- ❘
- C
- _
- )
- +
- r
- *
- r
- wherein t is an n-gram, C is a target category, C is a non-target category, and r is a parameter that favors the n-gram based on a frequency and distinctness of the n-gram;
- accessing at least a portion of the content;
- providing at least the portion of the content to a content classifier, wherein the content classifier is configured to determine a second classification score based on at least the portion of the content, the second classification score being separate from the first classification score;
- combining the first classification score with the second classification score to generate a combined classification score;
- comparing the combined classification score to a threshold;
- and
- classifying the content into the target category or the non-target category based on the comparison of the classification score to the threshold.
32 more elements
Dependencies are inferred from the claim text.