Automatic identification of abstract online groups

• US 8,700,629 B2
• Filed: 07/03/2012
• Issued: 04/15/2014
• Est. Priority Date: 02/28/2011
• Status: Expired due to Fees

• Current Assignee: Battelle Memorial Institute
• Original Assignee: Battelle Memorial Institute
• Inventors: Engel, David W., Gregory, Michelle L., Bell, Eric B., Piatt, Andrew W., Cowell, Andrew J.
• Primary Examiner(s): Vy, Hung T
• Application Number:
  US13/540,759
• Publication Number:
  US 20130191390A1
• Time in Patent Office: 651 Days
• Field of Search: 707/737, 707/736, 707/738, 707/758
• US Class Current: 707/737
• CPC Class Codes: G06F 16/35 Clustering; Classification
• CPC Class Codes: G06Q 10/10 Office automation; Time management
• CPC Class Codes: G06Q 50/01 Social networking

Abstract

Online abstract groups, in which members aren'"'"'t explicitly connected, can be automatically identified by computer-implemented methods. The methods involve harvesting records from social media and extracting content-based and structure-based features from each record. Each record includes a social-media posting and is associated with one or more entities. Each feature is stored on a data storage device and includes a computer-readable representation of an attribute of one or more records. The methods further involve grouping records into record groups according to the features of each record. Further still the methods involve calculating an n-dimensional surface representing each record group and defining an outlier as a record having feature-based distances measured from every n-dimensional surface that exceed a threshold value. Each of the n-dimensional surfaces is described by a footprint that characterizes the respective record group as an online abstract group.

Claims

1. A computer-implemented method of automatically identifying online abstract groups comprising entities that exhibit shared interests and/or characteristics, the method comprising:

Harvesting records from social media, each record comprising a social-media posting and being associated with one or more entities;

Extracting content-based and structure-based features from each record, each feature stored on a data storage device and comprising a computer-readable representation of an attribute of one or more records;

Grouping records into record groups according to the features of each record using clustering, classifying, and/or filtering algorithms executed by one or more processors;

Calculating an n-dimensional surface representing each record group, each n-dimensional surface described by a footprint that characterizes the respective record group as an online abstract group;

Defining an outlier as a record having feature-based distances measured from every n-dimensional surface that exceed a threshold value.

2. The computer-implemented method of claim 1, wherein at least a portion of the records in an online abstract group lacks explicit connections between records, between entities, and/or between records and entities.

3. The computer-implemented method of claim 1, wherein the social media posting comprises a blog post, at least one comment, or both.

4. The computer-implemented method of claim 3, wherein a structure-based feature comprises a duration of time from an initial blog post to a final comment regarding the initial blog post.

5. The computer-implemented method of claim 3, wherein a structure-based feature comprises lag time between blog posts, between comments, or between blog posts and comments.

6. The computer-implemented method of claim 1, wherein the social media posting comprises a tweet.

7. The computer-implemented method of claim 1, wherein the records comprise one or more foreign texts.

8. The computer-implemented method of claim 1, further comprising extracting entity-based features from entities associated with records.

9. The computer-implemented method of claim 8, wherein said extracting entity-based features comprises utilizing one or more link analysis techniques.

10. The computer-implemented method of claim 1, wherein the automatically identifying abstract online groups does not utilize link analysis techniques.

11. The computer-implemented method of claim 1, further comprising calculating an affect score as a content-based feature of at least one record.

12. The computer-implemented method of claim 1, wherein a content-based feature comprises a classification of degree of agreement/disagreement.

13. The computer-implemented method of claim 1, wherein a content-based feature comprises keywords of records, keywords within records, or both.

14. The computer-implemented method of claim 1, wherein a structure-based feature comprises a duration of time from initiation to conclusion of a record.

15. The computer-implemented method of claim 1, wherein a structure-based feature comprises a lag time between entries within in a record, a lag time between records, or both.

16. The computer-implemented method of claim 1, wherein the features are derived from statistical analysis on the representation of one or more attributes of one or more records.

17. The computer-implemented method of claim 1, further comprising presenting a visual representation that depicts the n-dimensional surface on a display device.

18. The computer-implemented method of claim 1, wherein at least one content-based or structure-based feature characterizes culturally defined behaviors within and/or between online abstract groups.