The fastest and easiest way to implement speech-to-text conversion for a small subset of speech

I want to implement a system that accepts speech through a microphone on my Mac OS x. I know that arbitrary speech recognition is almost impossible without training, so I want to limit it to 10 simple sentences. It must recognize with a high degree of accuracy which of these 10 sentences are speaking, generate text, and add a record to the remote MySQL database.

If this is a system architecture that I want to implement, can someone give me an overview of what would be the best way to implement this system? I am looking for ideas such as open source libraries to keep coding to a minimum as this is just a prototype application to demonstrate. I am mainly looking for a quick and easy solution. Thanks!