Use of Naive Bayes along with SVMs

I plan to use Naive Bayes in the following manner:

• As a pre-processing step for SVMs:
  
• Feature Selection - reduce the number of terms drastically to increase SVM scalability
• Enhancing the training set - increase the size of it
  
• As part of an ensemble of techniques - implement 3 techniques (Multivariate, Multinomial and SVM) and use the results from all three to build confidence in the classification