Analysing Dependency Dynamics in Web Data

resource thumbnail

Remove from Bookmarks

Do you really want to remove?
This action cannot be undone. Choose 'Cancel' to stop and go back.
Cancel
Ratings: 0
  • Which text to add here??

Added by benbanbun on 2010-11-09 11:20

» Viewed 614 times
» Favorited by 0 user(s)
» 0 Comments
» This resource has public visibility

Holder of Rights:

License: unknown

Creator(s): Felix Bießmann, Andreas Harth

Description:
Modern web sites provide easy access to large amounts of
data via open application programming interfaces. Users interacting with these sites constantly change the underlying data sets, which can be represented in graph-structured form. Nodes in these dynamic graph structures exhibit dependencies over time (for example, one node changes before other nodes change in the same way). Analysing these dependencies is crucial for understanding and predicting the dynamics inherent to temporally changing graph structures on the web. When the graphs become large however, it is not feasible to take into account all properties of the graph and in general it is unclear how to choose the appropriate features. Moreover, comparing two nodes becomes difficult, if the nodes do not share exactly the same features. In this work we propose an algorithm that automatically learns the features that govern temporal dependencies between nodes in large dynamic graph structures. We present preliminary results of applying the algorithm to data collected from the web, discuss potential extensions of the framework and anticipate how a major problem in machine learning, sparse data, could be tackled by leveraging Linked Data.

Add to Collection

You don't have any collections yet. Click here to create your first collection!

Okay

Share to Group

You don't have any group you can share this resource with: the resource is already shared to all groups you are member in. Click here to see available groups!

Okay

Create QR Code

Please select the URI for the QR Code:




Create Cancel

Tags

sort: alphabetically • by frequency
use blanks to separate tags

Comments

Analysing Dependency Dynamics in Web Data Modern web sites provide easy access to large amounts of data via open application programming interfaces. Users interacting with these sites constantly change the underlying data sets, which can be represented in graph-structured form. Nodes in these dynamic graph structures exhibit dependencies over time (for example, one node changes before other nodes change in the same way). Analysing these dependencies is crucial for understanding and predicting the dynamics inherent to temporally changing graph structures on the web. When the graphs become large however, it is not feasible to take into account all properties of the graph and in general it is unclear how to choose the appropriate features. Moreover, comparing two nodes becomes difficult, if the nodes do not share exactly the same features. In this work we propose an algorithm that automatically learns the features that govern temporal dependencies between nodes in large dynamic graph structures. We present preliminary results of applying the algorithm to data collected from the web, discuss potential extensions of the framework and anticipate how a major problem in machine learning, sparse data, could be tackled by leveraging Linked Data. Felix Bießmann, Andreas Harth