System and method for managing user attention by detecting hot and cold topics in social indexes

• US 8,239,397 B2
• Filed: 01/27/2009
• Issued: 08/07/2012
• Est. Priority Date: 01/27/2009
• Status: Active Grant

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Palo Alto Research Center, Inc. (Xerox Holdings Corp.)
• Inventors: Good, Lance E., Stefik, Mark J., Mittal, Sanjay
• Primary Examiner(s): SYED, FARHAN M
• Application Number:
  US12/360,834
• Publication Number:
  US 20100191742A1
• Time in Patent Office: 1,288 Days
• Field of Search: None
• US Class Current: 707/758
• CPC Class Codes: G06F 16/35 Clustering; Classification
• CPC Class Codes: G06F 16/353 into predefined classes
• CPC Class Codes: G06F 16/355 Class or cluster creation or modification
• CPC Class Codes: G06F 16/358 Browsing; Visualisation therefor
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

A system and method for managing user attention by detecting hot topics in social indexes is provided. Articles of digital information and at least one social index are maintained. The social index includes topics that each relate to one or more of the articles. Topic models matched to the digital information are retrieved for each topic. The articles are classified under the topics using the topic models. Each of the topics in the social index is evaluated for hotness. A plurality of time periods projected from the present is defined. Counts of the articles appearing under each time period are evaluated. The topics exhibiting a rising curve in the count of the articles that increases with recency during the time periods are chosen. Quality of the articles within the topics chosen is analyzed. The topics including the articles having acceptable quality are presented.

Claims

1. A computer-implemented method for managing user attention by detecting hot topics in social indexes, comprising the steps of:

maintaining articles of digital information and at least one social index comprising topics that each relate to one or more of the articles;

retrieving topic models matched to the digital information for each topic, comprising;

determining a form of training used to form the topic models;

generating a training score under example-based training for each topic chosen, comprising;

counting numbers of positive and negative training examples;

setting a training score in proportion to the positive and negative training examples numbers used;

identifying a number of characteristic words comprised in the articles relating to the topic; and

adjusting the training score in proportion to the number of the characteristic words; and

favoring those topics with the higher training scores, wherein the topic models comprise at least one of fine-grained topic models and coarse-grained topic models;

classifying the articles under the topics using the topic models;

evaluating each of the topics in the social index for hotness, comprising;

defining a plurality of time periods projected from the present;

evaluating counts of the articles appearing under each time period;

choosing the topics exhibiting a rising curve in the count of the articles that increases with recency during the time periods; and

analyzing quality of the articles within the topics chosen; and

presenting the topics comprised of the articles having acceptable quality,

wherein the steps are performed on a suitably-programmed computer.

2. A method according to claim 1, wherein the time periods comprise, in order of decreasing weight, one or more of day over week, day over month, and week over month.

3. A method according to claim 1, further comprising the steps of:

for each of the topics chosen, determining closeness of the topic to a center of the topic; and

favoring those topics comprised of the articles close to the center of the topic.

4. A method according to claim 1, further comprising:

determining user metrics for the articles counted, comprising the steps of;

counting a number of times that each article appearing under each time period has been read; and

finding a number of votes on each article appearing under each time period has been read; and

favoring those topics comprised of the articles with the stronger user metrics.

5. A method according to claim 1, further comprising the steps of:

setting a candidate score for each of the topics based on an increase in the count of the articles under each time period;

discounting each candidate score for those topics comprised of a high article count percentile rise and low article count;

discounting each candidate score for those topics comprised of a roll-up topic;

enhancing each candidate score comprised of popular articles; and

favoring those topics with the higher candidate scores.

6. A method according to claim 1, wherein the quality of the articles comprises one or more of recent growth during the time periods;

source reputation, and increase in number of sources.

7. A computer-implemented method for managing user attention by detecting cold topics in social indexes, comprising the steps of:

maintaining articles of digital information and at least one social index comprising topics that each relate to one or more of the articles, wherein each article includes each of content and an indication of recency;

retrieving topic models matched to the digital information for each topic, comprising;

determining a form of training used to form the topic models;

generating a training score under example-based training for each topic chosen, comprising;

counting numbers of positive and negative training examples;

setting a training score in proportion to the positive and negative training examples numbers used;

identifying a number of characteristic words comprised in the articles relating to the topic; and

adjusting the training score in proportion to the number of the characteristic words; and

favoring those topics with the higher training scores, wherein the topic models comprise at least one of fine-grained topic models and coarse-grained topic models;

classifying the articles under the topics using the topic models;

evaluating each of the topics in the social index for coldness, comprising;

maintaining a coldness threshold time period; and

eliding those articles comprised of the indication of recency that exceeds the coldness threshold time period; and

presenting the topics that were not elided,

wherein the steps are performed on a suitably-programmed computer.

8. A computer-implemented system for managing user attention by detecting hot topics in social indexes, comprising:

an electronic database, comprising;

articles of digital information and at least one social index comprising topics that each relate to one or more of the articles maintained for social indexing; and

topic models matched to the digital information for each topic;

a computer coupled with the electronic database and comprising a processor and memory within which code for execution by the processor is stored, further comprising;

a training evaluation module configured to determine a form of training used to form the topic models;

a training scoring module configured to generate a training score under example-based training for each topic chosen, comprising;

a counting module configured to count numbers of positive and negative training examples, to set a training score in proportion to the positive and negative training examples numbers used;

a characteristic word module configured to identify a number of characteristic words comprised in the articles relating to the topic; and

an adjustment module configured to adjust the training score in proportion to the number of the characteristic words,

wherein those topics with the higher training scores are favored and the topic models comprise at least one of fine-grained topic models and coarse-grained topic models;

a classifier module configured to classify the articles under the topics using the topic models;

an evaluation module configured to evaluate each of the topics in the social index for hotness, comprising;

a plurality of electronically-stored time periods projected from the present;

an article count evaluation module configured to evaluate counts of the articles appearing under each time period;

a topic chooser module configured to choose the topics exhibiting a rising curve in the count of the articles that increases with recency during the time periods; and

an analysis module configured to analyze quality of the articles within the topics chosen; and

a user interface configured to visually present the topics comprised of the articles having acceptable quality.

9. A system according to claim 8, wherein the time periods comprise, in order of decreasing weight, one or more of day over week, day over month, and week over month.

10. A system according to claim 8, wherein the processor and memory further comprise:

an topic evaluation module configured to determine, for each of the topics chosen, closeness of the topic to a center of the topic, and to favor those topics comprised of the articles close to the center of the topic.

11. A system according to claim 8, wherein the processor and memory further comprise:

a user metrics evaluation module configured to determine user metrics for the articles counted, comprising

a read count module configured to count a number of times that each article appearing under each time period has been read; and

a vote count module configured to find a number of votes on each article appearing under each time period has been read,

wherein those topics comprised of the articles with the stronger user metrics are favored.

12. A system according to claim 8, wherein the processor and memory further comprise:

a candidate scoring module configured to set a candidate score for each of the topics based on an increase in the count of the articles under each time period, to discount each candidate score for those topics comprised of a high article count percentile rise and low article count, to discount each candidate score for those topics comprised of a roll-up topic, to enhance each candidate score comprised of popular articles, and to favor those topics with the higher candidate scores.

13. A system according to claim 8, wherein the quality of the articles comprises one or more of recent growth during the time periods;

source reputation, and increase in number of sources.

14. A computer-implemented system for managing user attention by detecting cold topics in social indexes, comprising:

an electronic database, comprising;

articles of digital information and at least one social index comprising topics that each relate to one or more of the articles maintained for social indexing, wherein each article includes each of content and an indication of recency; and

topic models matched to the digital information for each topic;

a computer coupled with the electronic database and comprising a processor and memory within which code for execution by the processor is stored, further comprising;

a training determination module configured to determine a form of training used to form the topic models;

a training scoring module configured to generate a training score under example-based training for each topic chosen, comprising;

a counting module configured to count numbers of positive and negative training examples, to set a training score in proportion to the positive and negative training examples numbers used;

a characteristic word module configured to identify a number of characteristic words comprised in the articles relating to the topic; and

an adjustment module configured to adjust the training score in proportion to the number of the characteristic words,

wherein those topics with the higher training scores are favored and the topic models comprise at least one of fine-grained topic models and coarse-grained topic models;

a classifier module configured to classify the articles under the topics using the topic models;

an evaluation module configured to evaluate each of the topics in the social index for coldness, comprising;

an electronically-stored coldness threshold time period; and

an elision module configured to elide those articles comprised of the indication of recency that exceeds the coldness threshold time period; and

a user interface configured to visually present the topics that were not elided