# US 8,788,503 B1 · Content identification > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 8,788,503 B1 - Document ID: US8788503B1 - Status: Granted - Kind: B1 - Application number: 14/036,037 - Filed: 2013-09-25 - Granted: 2014-07-22 - Earliest claimed: 2007-10-17 - Assignees: Google Inc. - Applicants: Google Inc. - Inventors: Jay Yagnik - Web page: https://patentgrep.com/patent/US8788503B1 ## CPC - G06F 16/583: using metadata automatically derived from the content - G06F 18/231: Hierarchical techniques, i.e. dividing or merging pattern sets so as to obtain a dendrogram - G06V 10/7625: {Hierarchical techniques, i.e. dividing or merging patterns to obtain a tree-like representation; Dendograms} - G06V 20/35: {Categorising the entire scene, e.g. birthday party or wedding scene} - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## Abstract Systems, computer program products, and methods can identify a training set of content, and generate one or more clusters from the training set of content, where each of the one or more clusters represent similar features of the training set of content. The one or more clusters can be used to generate a classifier. New content is identified and the classifier is used to associate at least one label with the new content. ## Claims ### Claim 1 A method, comprising: generating, by a device comprising a processor, a plurality of distinct clusters from training content, wherein the plurality of clusters represent features of content items in the training content; identifying one or more conjunctions of the plurality of distinct clusters based on a respective probability of observing a feature of a cluster of the one or more conjunctions in a collection of the content items; scoring an identified one or more conjunctions based on a conditional probability that the identified one or more conjunctions is associated with a label; selecting, as a current conjunction, one of the scored identified one or more conjunctions that has a score that meets a defined condition; and generating one or more higher-order child conjunctions for the current conjunction, wherein at least one of the one or more higher-order child conjunctions is a conjunction of the conjoined clusters of the current conjunction with one or more additional clusters not included in the conjoined clusters, and wherein the generating the one or more higher-order child conjunctions is performed if a stopping condition has not been reached. ### Claim 2 (depends on 1) The method of claim 1, further comprising: scoring at least one of the one or more of the higher-order child conjunctions based on a conditional probability that the at least one of the one or more higher-order child conjunction is associated with the label, wherein the scoring the at least one of the one or more of the higher-order child conjunctions is performed if a stopping condition has not been reached. ### Claim 3 (depends on 1) The method of claim 1, further comprising: determining that the stopping condition has been reached based on determining that a highest scoring one of the one or more higher-order child conjunctions has a score that is less than a score of the current conjunction; designating the highest scoring one of the one or more higher-order child conjunctions as the current conjunction based on determining that the stopping condition has not been reached; and generating a classifier for the label from the current conjunction. ### Claim 4 (depends on 1) The method of claim 1, wherein the identifying the one or more conjunctions of the plurality of distinct clusters, further comprises generating a distributed data set that indicates probabilities of observing the features in the collection of content items. ### Claim 5 (depends on 1) The method of claim 1, wherein the content items comprise images. ### Claim 6 (depends on 1) The method of claim 1, wherein the content items comprise text. ### Claim 7 (depends on 1) The method of claim 1, wherein the conjunction comprises a set of two or more of the plurality of distinct clusters. ### Claim 8 (depends on 1) The method of claim 1, wherein the cluster represents a collection of content item portions having features with a defined level of similarity. ### Claim 9 (depends on 1) The method of claim 1, wherein the content items comprise images, the method further comprising: examining the images at one or more scales or positions; and categorizing, into clusters, portions of the images having a defined level of similarity. ### Claim 10 (depends on 1) The method of claim 1, wherein the content items are parsed into regions, the method further comprising: grouping respective ones of the regions having patterns with a defined level of similarity into respective clusters. ### Claim 11 A computer-readable storage medium storing instructions that, in response to execution, cause a device comprising a processor to perform operations, comprising: generating a plurality of distinct clusters from training content, wherein the plurality of clusters represent features of content items in the training content; identifying one or more conjunctions of the plurality of distinct clusters based on a respective probability of observing a feature of a cluster of the one or more conjunctions in a collection of the content items; scoring an identified one or more conjunctions based on a conditional probability that the identified one or more conjunctions is associated with a label; selecting, as a current conjunction, one of the scored identified one or more conjunctions that has a score that meets a defined condition; and generating one or more higher-order child conjunctions for the current conjunction, wherein at least one of the one or more higher-order child conjunctions is a conjunction of the conjoined clusters of the current conjunction with one or more additional clusters not included in the conjoined clusters, and wherein the generating the one or more higher-order child conjunctions is performed if a stopping condition has not been reached. ### Claim 12 (depends on 11) The computer readable storage medium of claim 11, wherein the operations further comprise: scoring at least one of the one or more of the higher-order child conjunctions based on a conditional probability that the at least one of the one or more higher-order child conjunction is associated with the label, wherein the scoring the at least one of the one or more of the higher-order child conjunctions is performed if a stopping condition has not been reached. ### Claim 13 (depends on 11) The computer readable storage medium of claim 11, wherein the operations further comprise: determining that the stopping condition has been reached based on determining that a highest scoring one of the one or more higher-order child conjunctions has a score that is less than a score of the current conjunction; designating the highest scoring one of the one or more higher-order child conjunctions as the current conjunction based on determining that the stopping condition has not been reached; and generating a classifier for the label from the current conjunction. ### Claim 14 (depends on 11) The computer readable storage medium of claim 11, wherein the identifying the one or more conjunctions of the plurality of distinct clusters further comprises generating a distributed data set that indicates probabilities of observing the features in the collection of content items. ### Claim 15 (depends on 11) The computer readable storage medium of claim 11, wherein the content items comprise images. ### Claim 16 (depends on 11) The computer readable storage medium of claim 11, wherein the content items comprise text. ### Claim 17 (depends on 11) The computer readable storage medium of claim 11, wherein the conjunction comprises a set of two or more of the plurality of distinct clusters. ### Claim 18 (depends on 11) The computer readable storage medium of claim 11, wherein the cluster represents a collection of content item portions having features with a defined level of similarity. ### Claim 19 (depends on 11) The computer readable storage medium of claim 11, wherein the content items comprise images, and wherein the operations further comprise: examining the images at one or more scales or positions; and categorizing, into clusters, portions of the images having a defined level of similarity. ### Claim 20 (depends on 11) The computer readable storage medium of claim 11, wherein the content items are parsed into regions, and wherein the operations further comprise: grouping respective ones of the regions having patterns with a defined level of similarity into respective clusters. ### Claim 21 A system, comprising: a processor, communicatively coupled to a memory, that executes or facilitates execution of instructions to at least: generate a plurality of distinct clusters from training content, wherein the plurality of clusters represent features of content items in the training content; identify one or more conjunctions of the plurality of distinct clusters based on a respective probability of observing a feature of a cluster of the one or more conjunctions in a collection of the content items; score an identified one or more conjunctions based on a conditional probability that the identified one or more conjunctions is associated with a label; select, as a current conjunction, one of the scored identified one or more conjunctions that has a score that meets a defined condition; and generate one or more higher-order child conjunctions for the current conjunction, wherein at least one of the one or more higher-order child conjunctions is a conjunction of the conjoined clusters of the current conjunction with one or more additional clusters not included in the conjoined clusters, and wherein the generating the one or more higher-order child conjunctions is performed if a stopping condition has not been reached. ### Claim 22 (depends on 21) The system of claim 21, wherein the processor further executes or facilitates execution of instructions to at least: score at least one of the one or more of the higher-order child conjunctions based on a conditional probability that the at least one of the one or more higher-order child conjunction is associated with the label, wherein the scoring the at least one of the one or more of the higher-order child conjunctions is performed if a stopping condition has not been reached. ### Claim 23 (depends on 21) The system of claim 21, wherein the processor further executes or facilitates execution of instructions to at least: determine that the stopping condition has been reached based on determining that a highest scoring one of the one or more higher-order child conjunctions has a score that is less than a score of the current conjunction; designate the highest scoring one of the one or more higher-order child conjunctions as the current conjunction based on determining that the stopping condition has not been reached; and generate a classifier for the label from the current conjunction. ### Claim 24 (depends on 21) The system of claim 21, wherein identifying the one or more conjunctions of the plurality of distinct clusters further comprises generation of a distributed data set that indicates probabilities of observing the features in the collection of content items. ### Claim 25 (depends on 21) The system of claim 21, wherein the content items comprise images. ### Claim 26 (depends on 21) The system of claim 21, wherein the content items comprise text. ### Claim 27 (depends on 21) The system of claim 21, wherein the conjunction comprises a set of two or more of the plurality of distinct clusters. ### Claim 28 (depends on 21) The system of claim 21, wherein the cluster represents a collection of content item portions having features with a defined level of similarity. ### Claim 29 (depends on 21) The system of claim 21, wherein the content items comprise images, and wherein the processor further executes or facilitates execution of instructions to at least: examine the images at one or more scales or positions; and categorize, into clusters, portions of the images having a defined level of similarity. ### Claim 30 (depends on 21) The system of claim 21, wherein the content items are parsed into regions, and wherein the processor further executes or facilitates execution of instructions to at least: group respective ones of the regions having patterns with a defined level of similarity into respective clusters.