I think you might need to rephrase the question for better understanding. On a general note, classifiers (be it SVM or any other) works on your dataset. So, the first thing would be to get your News dataset (assuming you are referring to News data) well formatted. Then SVM or any other applicable classifier can then be used to extract patterns from the respective classes in the dataset.

To deal with this;

#First Data has to be converted into operable form

if the news item is still in a string/text form or some form of categorical data; you have to convert these to numerical value

#identify some number of relevant features you can extract from your news item and for each news item, if that feature is present you can give it a value of 1 else 0 or even in between

#Create classes( in this case its either IT recruitment [1] or not [0]; ie, binary)

#each observation or sample (ie. each row vector) should be assigned corresponding classes base on the results of the combination of those features.

You know which features need to be present for it to be IT recruitment, so decide on it

#All we have done so far is to convert data into appropriate form for training

NOW LETS CREATE OUR SVM MODEL

#SVM depends on three major parameters

1. kernel[ either 'rbf' or 'linear']

2.gamma: can take various values eg 0.01, 0.10, 0.001

3. C: C determines the position of your decision line; if its too big leads to overfitting .

DO NECESSARY IMPORTATIONS Eg:

from sklearn import svm

from sklearn.cross_validation import train_test_split

etc

X=data(features)

Y=response data[made up of classes for each row or sample or observation)

x_trn, x_tst, y_trn, y_tst=train_test_split(X,Y, test_size=0.5, random_state=4)

svc_clf=svm.SVC(kernel='rbf', decision_function_shape='ovo')

svc_clf.fit(x_trn,y_trn)

prediction=svc_clf.predict(x_tst)

YOU CAN ALSO CREATE A DICTIONARY

THERE ARE DOCUMENTS ON CONVERTION OF DATA TO NUMERICAL IF IT IS IN CATEGORICAL FORM

https://chrisalbon.com/python/convert_categorical_to_numeric.html

I Hope I Didn't Misconstrue Your Question. AND Wish this is helpful.

Thanks

There a couple of solutions, which would really reflect your problem in solving, but is;t your question is a little hard to understand, my suggestion, that sklearn lib of python, resources much have to offer in this regards..

Your question is a bit ambiguous, but "scikit-learn" package in python can solve your problem.