text mining - customizing Stanford-NER for detecting Actor, Director, production company names in movie reviews -

i want train model detect names of actor, director, production company movie reviews. have tokenized , tagged dataset now, have set feature properties. searched lot featurefactoryner didn't find reasonable way completly understand class. question how can define our own features in featurefactoryner in order train model in right way? please in understanding featurefactory, properties , defining customize features?