Department of Computing

COMP348 Document Processing and the Semantic Web

Practical, Week 9

Graphs and Folksonomies

The following graph is a subset of the first 2000 edges of the graph used in assignment 2. The graph represents (part of) a folksonomy where resources are tagged with user-defined tags.

• Download the graph

• Write a program that returns the 10 most popular tags.
• Two tags cooccur in a resource if they tag the same resource. Write a program that returns the 10 most common pair of tags that cooccur in a resource.
• The list of cooccurring tags of a tag t is the list of tags that share a resource with t. Write a program that returns the 10 tags with the longest list of cooccurring tags.

Fine-tuning a SVM Classifier (for those aiming at bonus marks in assignment 2)

SVM classifiers are very powerful because of their ability to hadle a wide range of classification tasks, but often the default parameters are not optimal and sometimes you get much better results by fine-tuning them. This basically means doing a systematic series of tests with selected parameters, narrowing down the options until you get the final parameters.

• Read this practical guide to SVM classification. It details a process to determine the parametres of LIBSVM.
• Install LIBSVM in your computer and apply what they say in the manual to optimise the parameters of the example heart_scale that comes with the LIBSVM software.

If you have done all the above exercises, work on exercises of past weeks.