Pretty much, yes. In saera_processing.py you should generally try to match keywords instead of the whole sentence, in case it isn't recognized perfectly. But you will have to regenerate the language model if you add words to the vocab. I used the online tool at http://www.speech.cs.cmu.edu/tools/lmtool.html to build the files manually; in the next release Saera will build it herself so that I can support other languages. You will also need to edit two numbers in saera.py to refer to the new .lm and .dic files the tool will generate.