training and testing speech files on kaldi speech recognition system

I have a speech corpus, divided into four sub-corpora (dialects). I need to extract the Word Error Rate (WER) for every sub-corpus. For every sub-corpus I need to do training and testing 18 times by changing "tied states" and "gaussian densities". Then, I want to extract the full-corpus WER in same manner. So, the total number of training and testing will be 18*5 times. The corpus is ready for using in CMU spninx, so I will attach all the files that are related to sphinx. These files need minor change to work on kaldi. To sum up: I need to (1) transform the project from sphinx to kaldi (2) do training and testing & extract WERs in your machine (3) provide me with the new project (kaldi) and the results (4) run at least one example in my machine.