SVM is mainly used for classification, I wonder how do u want to use it as extraction of important news. I will suggest follow these steps

Feature Extraction:

you may wish to use bag of word approach to extract the concepts from the news. u should use either python of java for this purpose. the steps included are:

1. Tokenization of the news reports (n grams) taking two or multiple of the words from the news.

2. stop word elimination

3. stemming

once it is done then use wordnet to generate the concepts.

after extracting the news concepts you wish to visualize them as clusters,. that would be helpful for important news extraction.

If you can write more about your problem I might give u more suggestions

Hi

SVM helping in feature extraction so first of all you set a word combination then make a pattern after that applying the SVM  for feature extraction each para

This is purely a web-scrapping task. You do not need SVM for this at all. As Nadeem suggested you can go for the Bag of Words model for meaningful extraction of your required data.

Cheers.