METHOD AND APPARATUS FOR PREDICTION OF COMMUNITY REACTION TO A POST

• US 20140088944A1
• Filed: 09/24/2012
• Published: 03/27/2014
• Est. Priority Date: 09/24/2012
• Status: Active Grant

• Current Assignee: Adobe Inc.
• Original Assignee: Adobe Systems Incorporated (Adobe Inc.)
• Inventors: Natarajan, Anandhavelu, Srinivasan, Balaji Vasan, Gupta, Vineet, Ganesan, Anand, Jain, Anuj, Revankar, Shriram, Singh, Japnik, Polineni, Bharat
• Granted Patent: US 9,852,239 B2
• Time in Patent Office: Days
• US Class Current: 703/13
• CPC Class Codes: G06F 30/20 Design optimisation, verification or simulation optimisation, verification or simulation of circuit designs G06F30/30
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 50/01 Social networking

Abstract

A method and apparatus for prediction of community reaction to a post for an online social community is disclosed. The method comprises receiving a proposed post as input to a generated prediction model prior to the proposed post being posted to an online social community; predicting a community reaction to the proposed post using the prediction model; and displaying the predication, wherein the prediction comprises a sentiment score and at least one of a number of responses, a number of responders to the post, a longevity of the post, or a half-life of the post.

Claims

1. A computer-implemented method comprising:

receiving a proposed post as input to a generated prediction model prior to the proposed post being posted to an online social community;

predicting a community reaction to the proposed post using the prediction model; and

displaying the predication, wherein the prediction comprises a sentiment score and at least one of a number of responses, a number of responders to the post, a longevity of the post, or a half-life of the post.

2. The method of claim 1, wherein generating the prediction model comprises:

gathering a plurality of posts from one or more online social communities, wherein each post in the plurality of posts comprises at least one response to the post;

calculating a plurality of performance indicator scores for each post in the plurality of posts, wherein the plurality of performance indicators comprise a sentiment score for the at least one response and at least one of a number of responses to the post, a number of responders to the post, a longevity of the post, or a half-life of the post;

de-biasing the calculated scores; and

building a prediction model using the de-biased scores.

3. The method of claim 1, wherein displaying the prediction shows a plurality of bins in varying shades of color indicating the confidence in the prediction of the corresponding bin and further shows the text of the post that contribute to the reaction colored to reflect positive or negative performance indicator with a font size indicating the extremity of the reaction.

4. The method of claim 2, further comprising retraining the prediction model at predefined intervals.

5. The method of claim 2, wherein de-biasing comprises calculating the mean (m) and standard deviation (s) of the scores for all responses to a post and normalizing the score (r) to determine a de-biased score r′

, where r′

=(r−

m)/s.

6. The method of claim 2, wherein building the prediction model comprises:

accessing a plurality of posts and a score for each post in the plurality of posts;

retrieving a plurality of terms from each post in the plurality of posts;

calculating a conditional probability for each term in the plurality of terms;

determining the terms with a conditional probability that maximize entropy; and

adding the determined terms to the prediction model.

7. The method of claim 1, wherein predicting a reaction comprises:

identifying a plurality of terms and bigrams from the proposed post;

determining a class for which a prediction is to be made for each identified term and bigram in the plurality of identified terms and bigrams; and

determining a probability that the term and bigram in the plurality of identified terms and bigrams will be in the class.

8. An apparatus for predicting a reaction to a post comprising:

a reaction processor comprising a reaction predictor for receiving a proposed post as input to a trained prediction model prior to the proposed post being posted to an online social community, predicting a community reaction to the proposed post using the prediction model, and displaying the prediction, wherein the prediction comprises a sentiment score and at least one of a number of responses, a number of responders to the post, a longevity of the post, or a half-life of the post.

9. The apparatus of claim 8, further comprising:

a model generator for gathering a plurality of posts from one or more online social communities, wherein each post in the plurality of posts comprises at least one response to the post, and wherein the model generator comprises;

a performance indicator estimator for calculating a plurality of performance indicator scores for the at least one response, wherein the plurality of performance indicators comprise at least one of a number of responses to the post, a number of responders to the post, a longevity of the post, a sentiment score for the at least one response or a half-life of the post;

a sentiment classifier for calculating a sentiment score for the at least one response;

a response de-biasing module for de-biasing the calculated scores; and

a prediction trainer for building a prediction model using the de-biased scores.

10. The apparatus of claim 8, wherein displaying the prediction shows a plurality of bins in varying shades of color indicating the confidence in the prediction of the corresponding bin and further shows the text of the post that contribute to the reaction colored to reflect positive or negative performance indicator with a font size indicating the extremity of the reaction.

11. The apparatus of claim 9, wherein de-biasing comprises calculating the mean (m) and standard deviation (s) of the scores for all responses to a post and normalizing the score (r) to determine a de-biased score r′

, where r′

=(r−

m)/s.

12. The apparatus of claim 9, wherein building the prediction model comprises:

accessing a plurality of posts and a score for each post in the plurality of posts;

retrieving a plurality of terms from each post in the plurality of posts;

calculating a conditional probability for each term in the plurality of terms;

determining the terms with a conditional probability that maximize entropy; and

adding the determined terms to the prediction model.

13. The apparatus of claim 8, wherein predicting a reaction comprises:

identifying a plurality of terms and bigrams from the proposed post;

determining a class for which a prediction is to be made for each identified term and bigram in the plurality of identified terms and bigrams; and

determining a probability that the term and bigram in the plurality of identified terms and bigrams will be in the class.

14. A non-transient computer readable medium for storing computer instructions that, when executed by at least one processor cause the at least one processor to perform a method for predicting community reaction to a post comprising:

receiving a proposed post as input to a generated prediction model prior to the proposed post being posted to an online social community;

predicting a community reaction to the proposed post using the prediction model; and

displaying the predication, wherein the prediction comprises a sentiment score and at least one of a number of responses, a number of responders to the post, a longevity of the post, or a half-life of the post.

15. The computer readable medium of claim 14, wherein generating the prediction model comprises:

gathering a plurality of posts from one or more online social communities, wherein each post in the plurality of posts comprises at least one response to the post;

calculating a plurality of performance indicator scores for each post in the plurality of posts, wherein the plurality of performance indicators comprise a sentiment score for the at least one response and at least one of a number of responses to the post, a number of responders to the post, a longevity of the post, or a half-life of the post;

de-biasing the calculated scores; and

building a prediction model using the de-biased scores.

16. The computer readable medium of claim 14, wherein displaying the prediction shows a plurality of bins in varying shades of color indicating the confidence in the prediction of the corresponding bin and further shows the text of the post that contribute to the reaction colored to reflect positive or negative performance indicator with a font size indicating the extremity of the reaction.

17. The computer readable medium of claim 15, further comprising retraining the prediction model at predefined intervals.

18. The computer readable medium of claim 15, wherein de-biasing comprises calculating the mean (m) and standard deviation (s) of the scores for all responses to a post and normalizing the score (r) to determine a de-biased score r′

, where r′

=(r−

m)/s.

19. The computer readable medium of claim 15, wherein building the prediction model comprises:

accessing a plurality of posts and a score for each post in the plurality of posts;

retrieving a plurality of terms from each post in the plurality of posts;

calculating a conditional probability for each term in the plurality of terms;

determining the terms with a conditional probability that maximize entropy; and

adding the determined terms to the prediction model.

20. The computer readable medium of claim 14, wherein predicting a reaction comprises:

identifying a plurality of terms and bigrams from the proposed post;

determining a class for which a prediction is to be made for each identified term and bigram in the plurality of identified terms and bigrams; and

determining a probability that the term and bigram in the plurality of identified terms and bigrams will be in the class.