Measuring semantic incongruity within text data

• US 10,176,260 B2
• Filed: 02/10/2015
• Issued: 01/08/2019
• Est. Priority Date: 02/12/2014
• Status: Active Grant

• Current Assignee: Regents of The University of Minnesota (University of Minnesota)
• Original Assignee: Regents of The University of Minnesota (University of Minnesota)
• Inventors: Mahapatra, Amogh, Srivastava, Nisheeth, Srivastava, Jaideep
• Primary Examiner(s): Gortayo, Dangelino N
• Application Number:
  US14/618,838
• Publication Number:
  US 20150227626A1
• Time in Patent Office: 1,428 Days
• Field of Search: 707706, 707708, 707728, 707737, 707739, 707748, 707749
• CPC Class Codes: G06F 16/221 Column-oriented storage; Management thereof
• CPC Class Codes: G06F 16/2228 Indexing structures
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06F 16/31 Indexing; Data structures therefor; Storage structures
• CPC Class Codes: G06F 16/3331 Query processing
• CPC Class Codes: G06F 16/3334 Selection or weighting of terms from queries, including natural language queries
• CPC Class Codes: G06F 16/3335 Syntactic pre-processing, e.g. stopword elimination, stemming
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 16/3347 using vector based model
• CPC Class Codes: G06F 16/345 Summarisation for human users
• CPC Class Codes: G06F 16/35 Clustering; Classification
• CPC Class Codes: G06F 16/36 Creation of semantic tools, e.g. ontology or thesauri
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

A computing device may determine, based at least in part on one or more semantic exemplars associated with a digital object, a level of incongruity for the digital object, and output an indication of the level of incongruity for the digital object. For instance, a system may provide a web application that can support hedonic web surfing. By modeling Internet users as active information foragers instead of random surfers, a system may obtain quantitative measures of digital objects that users may find psychologically stimulating. The system may utilize a quantitative measure of the conceptual incongruity of digital objects that may predict how interesting users will find an object.

Claims

1. A method comprising:

processing, by a computing device, textual data associated with a digital object to identify within the textual data one or more semantic exemplars for the digital object;

generating, by the computing device, a data structure that represents a semantic space of semantic diversity and semantic anomalousness for the digital object and specifies a respective measure of similarity between each pairwise combination of the semantic exemplars identified within the textual data;

computing, by the computing device and based at least in part on the data structure, a level of incongruity for the digital object that represents a predicted estimation of interest of the digital object to a user and from a function that quantifies joint presence of semantic diversity and semantic anomalousness; and

executing a programmatic action based on the computed level of incongruity for the digital object, wherein executing the programming action comprises outputting, by the computing device, an indication of the level of incongruity for the digital object.

2. The method of claim 1, wherein generating the data structure that represents the semantic space for the digital object comprises:

for each pairwise combination of the semantic exemplars, the pairwise combination including a first semantic exemplar and a second semantic exemplar, determining the respective measure of similarity between the pairwise combination by accessing an index generated based on a set of resource objects and computing, using the index, a respective co-occurrence, in the set of resource objects, of the first and second semantic exemplars, and

generating, based at least in part on the respective co-occurrence between each pairwise combination of the semantic exemplars, a similarity matrix for the digital object, the similarity matrix representing the semantic space,

wherein rows of the similarity matrix each represent a first instance of a respective semantic exemplar from the one or more semantic exemplars, wherein columns of the similarity matrix each represent a second instance of the respective semantic exemplar, and wherein each value within the similarity matrix indicates the respective measure of similarity between the semantic exemplars represented by a row and column that each corresponds to the value.

3. The method of claim 1, wherein computing the level of incongruity for the digital object comprises:

computing, based on the respective measure of similarity between each pairwise combination of the semantic exemplars, a level of semantic diversity for the digital object by identifying a minimum measure of similarity such that, for at least one semantic exemplar from the one or more semantic exemplars, the respective measure of similarity between each pairwise combination that includes the at least one semantic exemplar is less than or equal to the minimum measure of similarity;

computing, based on the respective measure of similarity between each pairwise combination of the semantic exemplars, a level of semantic anomalousness for the digital object by computing, for each of the one or more semantic exemplars, a respective divergence score and determining, based on the respective divergence score for each of the one or more semantic exemplars, a number of anomalous semantic exemplars; and

computing the level of incongruity for the digital object as a function of both the level of semantic diversity and the level of semantic anomalousness.

4. The method of claim 3, wherein computing the respective divergence score for each of the one or more semantic exemplars comprises computing, for a particular semantic exemplar, the sum of each respective measure of similarity between the particular semantic exemplar and each semantic exemplar from the one or more semantic exemplars, and

wherein determining the number of anomalous semantic exemplars comprises determining semantic exemplars that have a respective divergence score above an upper bound and semantic exemplars that have a respective divergence score below a lower bound, wherein the upper bound and the lower bound are each determined based on an inter-quartile range of the respective divergence score for each of the one or more semantic exemplars.

5. The method of claim 1, wherein determining the level of incongruity for the digital object further comprises determining, based on the data structure that represents the semantic space for the digital object, the presence of a loosely knit sense-cluster.

6. The method of claim 1, wherein determining the level of incongruity for the digital object further comprises determining, based on the data structure that represents the semantic space for the digital object, the presence of one or more anomalies in the semantic space.

7. The method of claim 1, further comprising:

receiving, by the computing device, a set of search terms having at least one search term;

executing a search to identify, based at least in part on the at least one search term, a set of digital objects relevant to the at least one search term, each of the set of digital objects having a computed relevance score, and each of the set of digital objects being associated with respective one or more semantic exemplars, wherein the digital object comprises one of the set of digital objects; and

processing each of the identified digital objects to determine, based at least in part on the respective one or more semantic exemplars for each of the set of digital objects, a respective level of incongruity for each of the set of digital objects.

8. The method of claim 7, wherein executing the search comprises:

sending, to a search engine, a search query that includes the at least one search term; and

responsive to sending the search query, receiving, from the search engine, the set of digital objects.

9. The method of claim 7, further comprising:

determining a respective summarization of each digital object from the set of digital objects, the respective summarization comprising respective one or more summary terms, wherein the respective one or more semantic exemplars associated with each digital object comprise the respective one or more summary terms; and

computing the level of incongruity for each of the digital objects based on the respective summarization of each of the digital objects.

10. The method of claim 7, wherein the respective one or more semantic exemplars for each of the set of digital objects comprises the at least one search term.

11. The method of claim 7, wherein the at least one search term comprises at least one first search term, wherein the set of digital objects comprises a first set of digital objects, and wherein outputting the indication of the respective level of incongruity for the at least one digital object comprises outputting a first indication of the respective level of incongruity for at least one first digital object from the first set of digital objects, the method further comprising:

receiving, by the computing device, an indication of input to select a specified digital object from the first set of digital objects;

determining, by the computing device and based at least in part on the at least one first search term and the specified digital object, at least one second search term;

updating the set of search terms to include the second search term and executing a second search to identify, based at least in part on the updated set of search terms, a second set of digital objects relevant to the at least one second search term, each of the second set of digital objects having a computed relevance score, each of the second set of digital objects being associated with respective one or more semantic exemplars; and

processing each of the identified digital objects to determine, based at least in part on the respective one or more semantic exemplars for each of the second set of digital objects, a respective level of incongruity for each of the second set of digital objects.

12. The method of claim 1, wherein outputting, by the computing device, an indication of the level of incongruity for the digital object comprises determining a set of one or more search results including the digital object and outputting the set of search results.

13. The method of claim 1, wherein the digital object comprises at least one of:

a digital communication, a webpage, or video data and wherein the one or more semantic exemplars comprise at least one of;

one or more words, or one or more tags.

14. The method of claim 1, further comprising determining at least one of the one or more semantic exemplars based on metadata associated with the digital object.

15. The method of claim 1, wherein outputting the indication of the computed level of incongruity comprises scoring, based on the level of incongruity, the digital object as one of a set of one or more search results and outputting the set of search results.

16. The method of claim 1, wherein outputting the indication of the computed level of incongruity comprises outputting a report or data for display.

17. The method of claim 1, wherein outputting the indication of the computed level of incongruity comprises outputting a set of suggested material for a user.

18. A computing device, comprising:

at least one processor; and

at least one module operable by the at least one processor to;

process textual data associated with a digital object to identify within the textual data one or more semantic exemplars for the digital object;

generate a data structure that represents a semantic space of semantic diversity and semantic anomalousness for the digital object and specifies a respective measure of similarity between each pairwise combination of the semantic exemplars identified within the textual data;

compute, based at least in part on the data structure, a level of incongruity for the digital object that represents a predicted estimation of interest of the digital object to a user and from a function that quantifies joint presence of semantic diversity and semantic anomalousness; and

execute a programmatic action based on the computed level of incongruity for the digital object, wherein the programming action comprises outputting, by the computing device, an indication of the level of incongruity for the digital object.

19. The computing device of claim 18, wherein the at least one module operable to generate the data structure that represents the semantic space for the digital object is operable by the at least one processor to:

for each pairwise combination of the semantic exemplars, the pairwise combination including a first semantic exemplar and a second semantic exemplar, determine the respective measure of similarity between the pairwise combination by accessing an index generated based on a set of resource objects and computing, using the index, a respective co-occurrence, in the set of resource objects, of the first and second semantic exemplars, and

generate, based at least in part on the respective co-occurrence between each pairwise combination of the semantic exemplars, a similarity matrix for the digital object, the similarity matrix representing the semantic space,

wherein rows of the similarity matrix each represent a first instance of a respective semantic exemplar from the one or more semantic exemplars, wherein columns of the similarity matrix each represent a second instance of the respective semantic exemplar, and wherein each value within the similarity matrix indicates the respective measure of similarity between the semantic exemplars represented by a row and column that each corresponds to the value.

20. The computing device of claim 18, wherein the at least one module operable to determine the level of incongruity for the digital object is operable by the at least one processor to:

compute, based on the respective measure of similarity between each pairwise combination of the semantic exemplars, a level of semantic diversity for the digital object by identifying a minimum measure of similarity such that, for at least one semantic exemplar from the one or more semantic exemplars, the respective measure of similarity between each pairwise combination that includes the at least one semantic exemplar is less than or equal to the minimum measure of similarity;

compute, based on the respective measure of similarity between each pairwise combination of the semantic exemplars, a level of semantic anomalousness for the digital object by computing, for each of the one or more semantic exemplars, a respective divergence score and determining, based on the respective divergence score for each of the one or more semantic exemplars, a number of anomalous semantic exemplars; and

compute the level of incongruity for the digital object as a function of both the level of semantic diversity and the level of semantic anomalousness.

21. The computing device of claim 20, wherein the at least one module operable to compute the respective divergence score for each of the one or more semantic exemplars is operable by the at least one processor to compute, for a particular semantic exemplar, the sum of each respective measure of similarity between the particular semantic exemplar and each semantic exemplar from the one or more semantic exemplars, and

wherein the at least one module operable to determine the number of anomalous semantic exemplars is operable by the at least one processor to determine semantic exemplars that have a respective divergence score above an upper bound and semantic exemplars that have a respective divergence score below a lower bound, wherein the upper bound and the lower bound are each determined based on an inter-quartile range of the respective divergence score for each of the one or more semantic exemplars.

22. The computing device of claim 18, wherein the at least one module is further operable by the at least one processor to:

receive a set of search terms having at least one search term;

execute a search to identify, based at least in part on the at least one search term, a set of digital objects relevant to the at least one search term, each of the set of digital objects having a computed relevance score, and each of the set of digital objects being associated with respective one or more semantic exemplars, wherein the digital object comprises one of the set of digital objects; and

process each of the identified digital objects to determine, based at least in part on the respective one or more semantic exemplars for each of the set of digital objects, a respective level of incongruity for each of the set of digital objects.

23. The computing device of claim 18, wherein outputting the indication of the computed level of incongruity comprises scoring, by the computing device based on the level of incongruity, the digital object as one of a set of one or more search results and outputting the set of search results.

24. The computing device of claim 18, wherein outputting the indication of the computed level of incongruity comprises outputting a report or data for display.

25. The computing device of claim 18, wherein outputting the indication of the computed level of incongruity comprises outputting a set of suggested material for a user.

26. A non-transitory computer-readable storage medium encoded with instructions that, when executed, cause at least one processor to:

process textual data associated with a digital object to identify within the textual data one or more semantic exemplars for the digital object;

generate a data structure that represents a semantic space of semantic diversity and semantic anomalousness for the digital object and specifies a respective measure of similarity between each pairwise combination of the semantic exemplars identified within the textual data;

compute, based at least in part on the data structure, a level of incongruity for the digital object that represents a predicted estimation of interest of the digital object to a user and from combination of a score of semantic diversity and a score of semantic anomalousness; and

execute a programmatic action based on the computed level of incongruity for the digital object, wherein the programming action comprises outputting an indication of the level of incongruity for the digital object.

27. The computer-readable storage medium of claim 26, wherein outputting the indication of the computed level of incongruity comprises scoring, based on the level of incongruity, the digital object as one of a set of one or more search results and outputting the set of search results.

28. The computer-readable storage medium of claim 26, wherein outputting the indication of the computed level of incongruity comprises outputting a report or data for display.

29. The computer-readable storage medium of claim 26, wherein outputting the indication of the computed level of incongruity comprises outputting a set of suggested material for a user.