Systems and methods for identifying semantically and visually related content

• US 9,984,310 B2
• Filed: 01/22/2016
• Issued: 05/29/2018
• Est. Priority Date: 01/23/2015
• Status: Active Grant

• Current Assignee: HighSpot, Inc.
• Original Assignee: HighSpot, Inc.
• Inventors: Hoffman, Raphael, Dire, Nate, Christensen, Erik, Sharp, Oliver, Wortendyke, David, Gellock, Scot, Wahbe, Robert
• Primary Examiner(s): Koziol, Stephen R
• Assistant Examiner(s): Azima, Shaghayegh
• Application Number:
  US15/004,693
• Publication Number:
  US 20160217343A1
• Time in Patent Office: 858 Days
• CPC Class Codes: G06F 18/22 Matching criteria, e.g. proximity measures
• CPC Class Codes: G06F 18/23 Clustering techniques
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06V 10/761 Proximity, similarity or dissimilarity measures
• CPC Class Codes: G06V 10/762 using clustering, e.g. of similar faces in social networks
• CPC Class Codes: G06V 30/414 Extracting the geometrical structure, e.g. layout tree; Block segmentation, e.g. bounding boxes for graphics or text
• CPC Class Codes: G06V 30/418 Document matching, e.g. of document images

Abstract

Systems and methods for identifying semantically and/or visually related information among a set of content items, such content items that include similar concepts or that have similar visual aspects, are disclosed. The disclosed techniques provide tools for identifying related information among various content items, such as text pages and documents, presentation slides and slide decks, etc. The disclosed techniques provide improved methods for searching among content items, organizing content items into categories, and pruning redundant content. Furthermore, the disclosed techniques provide improvements to computation of various metrics, including usage, performance, and impact metrics.

Claims

1. A method, performed by a computing system having a memory and a processor, the method comprising:

receiving an indication of a plurality of content items;

for each of the plurality of content items,

identifying at least one content element of the content item, and

for each identified content element of the content item,

extracting information for the content element,

computing a plurality of semantic feature values for the content element based at least in part on the extracted information,

computing a plurality of visual feature values for the content element based at least in part on the extracted information, and

storing the feature values computed, for the content element, based at least in part on the extracted information;

receiving an indication of a first set of content items, wherein each content item of the first set of content items is a member of the plurality of content items;

for each pair of content items from among the first set of content items,

applying a similarity function to the pair of content items to generate a similarity value, and

storing, in association with each content item of the pair of content items, the generated similarity value;

for each content item of the first set of content items,

identifying similar content items based at least in part on the generated similarity values, and

storing, in association with the content item, references to at least one of the identified similar content items.

2. The method of claim 1, wherein the received indication of the first set of content items comprises a link to at least one content item of the first set of content items.

3. The method of claim 1, wherein computing a plurality of semantic feature values for the content element based at least in part on the extracted information comprises applying a latent semantic analysis engine to at least a portion of text extracted from the content element.

4. The method of claim 1, wherein extracting information for the content element comprises generating a thumbnail image for the content element.

5. The method of claim 1, wherein identifying the set of similar content items for a first content item based at least in part on the generated similarity values comprises identifying a predetermined number of similar content items.

6. The method of claim 1, wherein identifying the set of similar content items for a first content item based at least in part on the generated similarity values comprises:

for each content item pair that includes the first content item and a content item other than the first content item,

retrieving the similarity value generated for the content item pair.

7. The method of claim 6, further comprising:

for each retrieved similarity value for the content item pairs that include the first content item and a content item other than the first content item,

determining whether the retrieved similarity value exceeds a predetermined threshold, and

in response to determining that the retrieved similarity value exceeds a predetermined threshold,

identifying the content item other than the first content item.

8. The method of claim 1, further comprising:

splitting a first content item of the plurality of content items into a plurality of content elements.

9. The method of claim 8, wherein the first content item is a slide presentation deck wherein splitting the first content item of the plurality of content items into a plurality of content elements comprises splitting the slide presentation deck into a plurality of slides, wherein a second content item is a word processing document, the method further comprising splitting the word processing document into a plurality of pages.

10. The method of claim 1, further comprising:

clustering the first set of content items based at least in part on the generated similarity values.

11. The method of claim 1, wherein applying a similarity function to a first pair of content items to generate a similarity value comprises:

for a first content item of the first pair of content items, determining a value for each of a plurality of attributes;

for a second content item of the first pair of content items, determining a value for each of a plurality of attributes;

calculating a distance between the first content item of the first pair of content items and the second content item of the first pair of content items according to the following equation

12. The method of claim 1, further comprising:

for each of a plurality of metrics,

for each content item of a first set of similar content items, retrieving, for the content item, a value for the metric, and

aggregating the values retrieved for the metric.

13. A system for analyzing content items, the system comprising:

at least one processor;

a component configured to receive an indication of a plurality of content items;

a component configured to, for each of the plurality of content items,

identify at least one content element of the content item, and

for each identified content element of the content item,

extract information for the content element,

compute a plurality of semantic feature values for the content element based at least in part on the extracted information,

compute a plurality of visual feature values for the content element based at least in part on the extracted information, and

store the feature values computed, for the content element, based at least in part on the extracted information;

a component configured to receive an indication of a first set of content items, wherein each content item of the first set of content items is a member of the plurality of content items;

a component configured to, for each pair of content items from among the first set of content items,

apply a similarity function to the pair of content items to generate a similarity value for the pair of content items, and

store, in association with each content item of the pair of content items, the generated similarity value; and

a component configured to, for each content item of the first set of content items,

identify a set of similar content items based at least in part on the generated similarity values, and

store, in association with the content item, references to at least one content item of the identified set of similar content items,

wherein each of the components comprises computer-executable instructions stored in a memory for execution by the at least one processor.

14. The system of claim 13, further comprising:

a component configured to store, in association with each content item of a first pair of content items, the similarity value generated for the first pair of content items in each of first metadata associated with a first content item of the first pair of content items and second metadata associated with a second content item of the first pair of content items.

15. The system of claim 13, further comprising:

a component configured to, for each content item of a first set of similar content items identified for a first content item, store an indication of the content item of the first set of similar content items in a lookup table, wherein the indication is referenced in the lookup table by an indication of the first content item.

16. The system of claim 13, wherein the component configured to receive an indication of a first set of content items receives a copy of at least one content item of the first set of content items.

17. The system of claim 13, further comprising:

means for ingesting content; and

means for assessing similarity of content items.

21. The system of claim 13, further comprising:

a component configured to split a first content item of the plurality of content items into a plurality of content elements.

22. The system of claim 13, further comprising:

a component configured to cluster the first set of content items based at least in part on the generated similarity values.

23. The system of claim 13, further comprising:

a component configured to, for each of a plurality of metrics,

for each content item of a first set of similar content items, retrieve, for the content item, a value for the metric, and

aggregate the values retrieved for the metric.

18. A computer-readable storage medium, that is not a transitory, propagating signal, storing instructions that, when executed by a computing system having a processor, cause the computing system to perform a method comprising:

receiving an indication of a plurality of content items; and

for each of the plurality of content items,

identifying at least one content element of the content item, and

for each identified content element of the content item,

extracting information for the content element,

computing a plurality of feature values for the content element based at least in part on the extracted information,

wherein computing the plurality of feature values for the content element based at least in part on the extracted information comprises computing a plurality of visual feature values for the content element based at least in part on the extracted information, and

wherein computing the plurality of feature values for the content element based at least in part on the of semantic feature values for the content element based at least in part on the extracted information, and

storing the feature values computed for the content element based at least in part on the extracted information; and

receiving an indication of a first set of content items, wherein each content item of the first set of content items is a member of the plurality of content items;

for each pair of content items from among the first set of content items,

applying a similarity function to the pair of content items to generate a similarity value, and

storing, in association with each content item of the pair of content items, the generated similarity value;

for each content item of the first set of content items,

identifying similar content items based at least in part on the generated similarity values, and

storing, in association with the content item references to at least one of the identified similar content items.

19. The computer-readable storage medium, that is not a transitory, propagating signal, of claim 18, the method further comprising:

generating a weight value for a first computed visual feature value; and

applying the generated weight value to the first computed visual feature value.

20. The computer-readable storage medium, that is not a transitory, propagating signal, of claim 18, the method further comprising:

generating a weight value for a first computed semantic feature value; and

applying the generated weight value to the first computed semantic feature value.