Difference between revisions of "Lerman et al www 2010"
(Created page with ' == Citation == Lerman, K., and Hogg, T. 2010. Using a model of social dynamics to predict popularity of online content. In Proc. 19th Int. World Wide Web Conference. == Online…') |
|||
| (3 intermediate revisions by the same user not shown) | |||
| Line 1: | Line 1: | ||
| − | |||
== Citation == | == Citation == | ||
| Line 9: | Line 8: | ||
== Summary == | == Summary == | ||
| − | This paper proposes and verifies a claim that by modelling the collective users of a social media site, we can predict the popularity of items, using users' early reaction to them ( [[AddressesProblem::Popularity Prediction]]). They try to solve three main problems in this paper: | + | This [[Category::paper]] proposes and verifies a claim that by modelling the collective users of a social media site, we can predict the popularity of items, using users' early reaction to them ( [[AddressesProblem::Popularity Prediction]]). They try to solve three main problems in this paper: |
* They verify the validity of their model by showing that it matches with previous observations. | * They verify the validity of their model by showing that it matches with previous observations. | ||
| − | + | The model's prediction about how social influence matters is correct for 95% stories in the dataset. | |
| + | |||
[[File:SvsRGraph.png]] | [[File:SvsRGraph.png]] | ||
| Line 18: | Line 18: | ||
* Estimate the story quality based by using the model of evolution of popularity. | * Estimate the story quality based by using the model of evolution of popularity. | ||
| − | + | Using the Stochastic model, explained in the background section, they predict 'r' that measures how interesting a story. | |
* Predict popularity by modeling the initial popularity of the stories. | * Predict popularity by modeling the initial popularity of the stories. | ||
| Line 25: | Line 25: | ||
== Dataset == | == Dataset == | ||
| − | The dataset is taken from May and June 2006 posts on Digg.com. | + | The dataset is taken from May and June 2006 posts on [http://digg.com/ Digg.com]. |
== Background == | == Background == | ||
| − | === Dynamic model of Social Voting == | + | === Dynamic model of Social Voting === |
| + | It uses a [[UsesMethod::Stochastic Process Modeling]] | ||
The model is based on the equation: | The model is based on the equation: | ||
| + | |||
[[File:eqn.png]] | [[File:eqn.png]] | ||
| + | |||
where <math>N_(vote)(t)</math> stands for Number of votes story has received by time t after it was submitted to Digg. | where <math>N_(vote)(t)</math> stands for Number of votes story has received by time t after it was submitted to Digg. | ||
the v's are the visibility factors that the story gained through either while on front page, upcoming page or through friends' votes.(more details in the dataset section) | the v's are the visibility factors that the story gained through either while on front page, upcoming page or through friends' votes.(more details in the dataset section) | ||
r measures how interesting the story is and it is estimated by minimizing the [[wiki:Root_Mean_Square RMS] difference between the obseverd votes and model predictions on the data. | r measures how interesting the story is and it is estimated by minimizing the [[wiki:Root_Mean_Square RMS] difference between the obseverd votes and model predictions on the data. | ||
| + | The following parameters are generated by the model. | ||
| + | |||
| + | [[File:params.png]] | ||
== Study Plan == | == Study Plan == | ||
| Line 42: | Line 48: | ||
* Background on the model in discussion : | * Background on the model in discussion : | ||
| − | + | * http://arxiv.org/abs/0806.1918 | |
| − | + | * [http://arxiv.org/abs/0904.0016 Stochastic Models of User-Contributory Web Sites] | |
Latest revision as of 20:09, 1 October 2012
Contents
Citation
Lerman, K., and Hogg, T. 2010. Using a model of social dynamics to predict popularity of online content. In Proc. 19th Int. World Wide Web Conference.
Online Version
http://www.isi.edu/~lerman/papers/wfp0788-lerman.pdf
Summary
This paper proposes and verifies a claim that by modelling the collective users of a social media site, we can predict the popularity of items, using users' early reaction to them ( Popularity Prediction). They try to solve three main problems in this paper:
- They verify the validity of their model by showing that it matches with previous observations.
The model's prediction about how social influence matters is correct for 95% stories in the dataset.
The important observation here is that better connected users (higher S) are more successful in getting their less interesting stories promoted to the front page than poorly connected users.
- Estimate the story quality based by using the model of evolution of popularity.
Using the Stochastic model, explained in the background section, they predict 'r' that measures how interesting a story.
- Predict popularity by modeling the initial popularity of the stories.
They claim that their system models both story's intrinsic quality and social influence and hence do better than just extrapolation
Dataset
The dataset is taken from May and June 2006 posts on Digg.com.
Background
Dynamic model of Social Voting
It uses a Stochastic Process Modeling The model is based on the equation:
where stands for Number of votes story has received by time t after it was submitted to Digg. the v's are the visibility factors that the story gained through either while on front page, upcoming page or through friends' votes.(more details in the dataset section) r measures how interesting the story is and it is estimated by minimizing the [[wiki:Root_Mean_Square RMS] difference between the obseverd votes and model predictions on the data.
The following parameters are generated by the model.
Study Plan
- Motivation for the paper: M. Salganik, P. Dodds, and D. Watts. Experimental study of inequality and unpredictability in an artificial cultural market. Science, 311:854, 2006 full text
- Background on the model in discussion :
- http://arxiv.org/abs/0806.1918
- Stochastic Models of User-Contributory Web Sites
Related Work
- D. M. Wilkinson. Strong regularities in online peer production. In EC ’08: Proc. of the 9th ACM conference on Electronic commerce, pages 302–309, New York, NY, USA,2008. ACM.
- K. Lerman and A. Galstyan. Analysis of social voting patterns on digg. In Proc. of the 1st ACM SIGCOMM Workshop on Online Social Networks, 2008.
