Hi, I'm doing a research on ASR/NLP applications to develop a prototype for my local language. I have already developed a prototype based on Julius and the ASR is grammar based. I wonder how application like Siri works. Are they performing grammar based recognition? I know that recognition without grammar can produce very poor results. But grammar model limits the number of possible sentences that can be recognized it seems. Any ideas or tips on what technologies/theories I should read about would be much appreciated.