see this:

http://kavita-ganesan.com/rouge-howto

ROUGE has numerous parameters, including word stemming, stop word removal, and n-gram size. Different settings work best for different summarisation tasks. This means that different parameters need to be tested for new tasks. It is recall-oriented metric. 

There is also a Java implementation of ROUGE and how to use it: http://kavita-ganesan.com/content/rouge-2.0-documentation

Here is the original perl implementation and instructions for how to use it: http://kavita-ganesan.com/rouge-howto